Publications

@article{zhang2026automatic,
  selected = {true},
  author  = {Zhang, Xiaoquan and Zhang, Mai and Cui, Lin and Tso, Fung Po and Zhang, Zhen and Deng, Yuhui},
  title   = {Automatic Model Compression and Quantized Deployment of Convolutional Neural Networks on Programmable Data Planes},
  journal = {IEEE Transactions on Networking},
  year    = {2026},
  doi     = {10.1109/TON.2026.3725437},
  note    = {Early access},
}

@article{qian2026dvrm,
  author  = {Qian, Mimi and Cui, Lin and Tso, Fung Po and Deng, Yuhui and Zhang, Zheng and Jia, Weijia},
  title   = {dVRM: Cross-Switch Memory Sharing and Self-Adaptive Allocation in Distributed Data Plane},
  journal = {IEEE Transactions on Computers},
  volume  = {75},
  number  = {7},
  pages   = {2382--2393},
  year    = {2026},
  doi     = {10.1109/TC.2026.3678701},
}

@article{chen2026faasorc,
  author  = {Chen, Chen and Nagel, Lars and Cui, Lin and Jia, Weijia and Tso, Fung Po},
  title   = {FaasOrc: A bi-level function scheduling and caching framework for serverless edge computing},
  journal = {Journal of Network and Computer Applications},
  volume  = {251},
  pages   = {104490},
  year    = {2026},
  doi     = {10.1016/j.jnca.2026.104490},
}

@inproceedings{zhang2026monic,
  selected = {true},
  author    = {Zhang, Xiaoquan and Liang, Bowen and Tso, Fung Po and Deng, Yuhui and Zhang, Zhen and Wei, Kaimin and Jia, Weijia and Cui, Lin},
  title     = {Monic: In-Network Mixture-of-Experts Inference on Programmable Data Planes},
  booktitle = {IEEE INFOCOM 2026 - IEEE Conference on Computer Communications},
  year      = {2026},
  doi       = {10.1109/INFOCOM59046.2026.11571404},
}

@inproceedings{chen2026sprint,
  selected = {true},
  author    = {Chen, Bingzhen and Lau, WaiMing and Zhang, Xiaoquan and Tso, Fung Po and Cui, Lin},
  title     = {SPRINT: Line-Rate In-band Network Telemetry Recovery for Application Optimization},
  booktitle = {IEEE INFOCOM 2026 - IEEE Conference on Computer Communications},
  year      = {2026},
  doi       = {10.1109/INFOCOM59046.2026.11571646},
}

@misc{jia2026unified,
  author        = {Jia, Zihan and Wang, Ze and Chen, Chen and Xiao, Ziren and Tso, Fung Po},
  title         = {Unified 5G-IoT Framework with CAMARA Gateways and SDN Federation},
  howpublished  = {arXiv preprint arXiv:2604.27589},
  year          = {2026},
  url           = {https://arxiv.org/abs/2604.27589},
}

@misc{chen2025ndn,
  author        = {Chen, Chen and Jia, Zihan and Wang, Ze and Cui, Lin and Tso, Fung Po},
  title         = {Design and Evaluation of an NDN-Based Network for Distributed Digital Twins},
  howpublished  = {arXiv preprint arXiv:2505.04326},
  year          = {2025},
  url           = {https://arxiv.org/abs/2505.04326},
}

@article{qian2025disploy,
  author  = {Qian, Mimi and Cui, Lin and Zhang, Xiaoquan and Tso, Fung Po and Deng, Yuhui and Li, Zhetao and Jia, Weijia},
  title   = {DisPLOY: Target-Constrained Distributed Deployment for Network Measurement Tasks on Data Plane},
  journal = {IEEE Transactions on Parallel and Distributed Systems},
  volume  = {36},
  number  = {8},
  pages   = {1608--1619},
  year    = {2025},
  doi     = {10.1109/TPDS.2025.3572246},
}

@article{zhang2025enhancing,
  author  = {Zhang, Xiaoquan and Cui, Lin and Lau, WaiMing and Tso, Fung Po and Deng, Yuhui and Jia, Weijia},
  title   = {Enhancing In-Network Computing Deployment via Collaboration Across Planes},
  journal = {IEEE Transactions on Computers},
  volume  = {74},
  number  = {11},
  pages   = {3805--3817},
  year    = {2025},
  doi     = {10.1109/TC.2025.3603730},
}

@article{qian2025flxvrm,
  author  = {Qian, Mimi and Cui, Lin and Tso, Fung Po and Deng, Yuhui and Zhang, Zhen and Jia, Weijia},
  title   = {FlxVRM: Enabling Online Configuring Memory via Virtualization on Programmable Data Plane},
  journal = {IEEE Transactions on Services Computing},
  volume  = {18},
  number  = {4},
  pages   = {2103--2114},
  year    = {2025},
  doi     = {10.1109/TSC.2025.3583131},
}

@article{zhang2025monte,
  author  = {Zhang, Xiaoquan and Cui, Lin and Tso, Fung Po and Deng, Yuhui and Li, Zhetao and Jia, Weijia},
  title   = {Monte: SFCs Migration Scheme in the Distributed Programmable Data Plane},
  journal = {IEEE Transactions on Parallel and Distributed Systems},
  volume  = {36},
  number  = {4},
  pages   = {633--644},
  year    = {2025},
  doi     = {10.1109/TPDS.2025.3532467},
}

@inproceedings{zhang2025planner,
  author    = {Zhang, Xiaoquan and Lau, WaiMing and Cui, Lin and Tso, Fung Po and Liang, Zhuoqian and Zhang, Zheng and Deng, Yuhui},
  title     = {Planner: A Generative Graph Learning Framework for Noisy and Dynamic In-band Network Telemetry},
  booktitle = {2025 IEEE 33rd International Conference on Network Protocols (ICNP)},
  pages     = {1--11},
  year      = {2025},
  doi       = {10.1109/ICNP65844.2025.11192325},
}

@inproceedings{zhang2025quark,
  author    = {Zhang, Mai and Cui, Lin and Zhang, Xiaoquan and Tso, Fung Po and Zhang, Zhen and Deng, Yuhui and Li, Zhetao},
  title     = {Quark: Implementing Convolutional Neural Networks Entirely on Programmable Data Plane},
  booktitle = {IEEE INFOCOM 2025 - IEEE Conference on Computer Communications},
  pages     = {1--10},
  year      = {2025},
  doi       = {10.1109/INFOCOM55648.2025.11044654},
}

@article{zhang2025reducing,
  author  = {Zhang, Yuxiang and Cui, Lin and Tso, Fung Po and Lei, Xiaolin},
  title   = {Reducing tail latency for multi-bottleneck in datacenter networks: A compound approach},
  journal = {Computer Networks},
  volume  = {257},
  pages   = {110931},
  year    = {2025},
  doi     = {10.1016/j.comnet.2024.110931},
}

@inproceedings{zhang2024carlo,
  author    = {Zhang, Xiaoquan and Cui, Lin and Lau, WaiMing and Tso, Fung Po and Deng, Yuhui and Jia, Weijia},
  title     = {Carlo: Cross-Plane Collaboration for Multiple In-network Computing Applications},
  booktitle = {IEEE INFOCOM 2024 - IEEE Conference on Computer Communications},
  pages     = {2528--2537},
  year      = {2024},
  doi       = {10.1109/INFOCOM52122.2024.10621111},
}

In recent years, vehicular networks have seen a proliferation of applications and services such as image tagging, lane detection, and speech recognition. Many of these applications rely on Deep Neural Networks (DNNs) and demand low-latency computation. To meet these requirements, Vehicular Edge Computing (VEC) has been introduced to augment the abundant computation capacity of vehicular networks to complement limited computation resources on vehicles. Nevertheless, offloading DNN tasks to MEC (Multi-access Edge Computing) servers effectively and efficiently remains a challenging topic due to the dynamic nature of vehicular mobility and varying loads on the servers. In this paper, we propose a novel and efficient distributed DNN Partitioning And Offloading (DPAO), leveraging the mobility of vehicles and the synergy between vehicle–edge and edge–edge computing. We exploit the variations in both computation time and output data size across different layers of DNN to make optimized decisions for accelerating DNN computations while reducing the transmission time of intermediate data. In the meantime, we dynamically partition and offload tasks between MEC servers based on their load differences. We have conducted extensive simulations and testbed experiments to demonstrate the effectiveness of DPAO. The evaluation results show that, compared to offloaded all tasks to MEC server, DPAO reduces the latency of DNN tasks by 2.4x. DPAO with queue reservation can further reduce the task average completion time by 10%.

@article{ZHENG2024110607,
title = {DNN acceleration in vehicle edge computing with mobility-awareness: A synergistic vehicle–edge and edge–edge framework},
journal = {Computer Networks},
volume = {251},
pages = {110607},
year = {2024},
issn = {1389-1286},
doi = {https://doi.org/10.1016/j.comnet.2024.110607},
url = {https://www.sciencedirect.com/science/article/pii/S1389128624004390},
author = {Yuxin Zheng and Lin Cui and Fung Po Tso and Zhetao Li and Weijia Jia},
keywords = {Vehicular edge computing, Deep neural networks, Task partitioning},
abstract = {In recent years, vehicular networks have seen a proliferation of applications and services such as image tagging, lane detection, and speech recognition. Many of these applications rely on Deep Neural Networks (DNNs) and demand low-latency computation. To meet these requirements, Vehicular Edge Computing (VEC) has been introduced to augment the abundant computation capacity of vehicular networks to complement limited computation resources on vehicles. Nevertheless, offloading DNN tasks to MEC (Multi-access Edge Computing) servers effectively and efficiently remains a challenging topic due to the dynamic nature of vehicular mobility and varying loads on the servers. In this paper, we propose a novel and efficient distributed DNN Partitioning And Offloading (DPAO), leveraging the mobility of vehicles and the synergy between vehicle–edge and edge–edge computing. We exploit the variations in both computation time and output data size across different layers of DNN to make optimized decisions for accelerating DNN computations while reducing the transmission time of intermediate data. In the meantime, we dynamically partition and offload tasks between MEC servers based on their load differences. We have conducted extensive simulations and testbed experiments to demonstrate the effectiveness of DPAO. The evaluation results show that, compared to offloaded all tasks to MEC server, DPAO reduces the latency of DNN tasks by 2.4x. DPAO with queue reservation can further reduce the task average completion time by 10%.},
}

@ARTICLE{10494953,
  author={Zhang, Xiaoquan and Cui, Lin and Tso, Fung Po and Li, Wenzhi and Jia, Weijia},
  journal={IEEE Communications Magazine}, 
  title={IN3: A Framework for In-Network Computation of Neural Networks in the Programmable Data Plane}, 
  year={2024},
  volume={62},
  number={4},
  pages={96-102},
  doi={https://doi.org/10.1109/MCOM.001.2300587},
  keywords={Neural networks;Application specific integrated circuits;Task analysis;Memory management;Control systems;Programmable control;Computational modeling;Telecommunication traffic;Data models},
}

@inproceedings{sofoluwe2024mitigating,
  author    = {Sofoluwe, Toyin and Dora, Shirin and Phillips, Iain and Tso, Fung Po},
  title     = {Mitigating Cyber Threats from the Edge},
  booktitle = {Advances in Computational Intelligence Systems (UKCI 2024), Advances in Intelligent Systems and Computing},
  publisher = {Springer},
  pages     = {279--290},
  year      = {2024},
  doi       = {10.1007/978-3-031-78857-4_21},
}

@article{qian2024ioffsetintitelemetry,
  author            = {Qian, M and Cui, L and Tso, FP and Deng, Y and Jia, W},
  journal           = {IEEE Transactions on Services Computing},
  pages             = {1--12},
  publisher         = {Institute of Electrical and Electronics Engineers (IEEE)},
  title             = {<i>OffsetINT</i>: Achieving High Accuracy and Low Bandwidth for In-Band Network Telemetry},
  url               = {http://dx.doi.org/10.1109/tsc.2023.3323697},
  year              = {2024},
  doi               = {https://doi.org/10.1109/tsc.2023.3323697},
  eissn             = {1939-1374},
  publicationstatus = {published},
}

@article{leesmith2024datahungry,
  author  = {Lee-Smith, Matthew and Ross, Tracy and Wilson, Garrath and Tso, Fung Po and Cavazzi, Stefano and Morley, Jeremy},
  title   = {The data hungry home: A post-anthropocentric and generative design framework},
  journal = {International Journal of Design},
  volume  = {18},
  number  = {1},
  year    = {2024},
  url     = {https://repository.lboro.ac.uk/articles/journal_contribution/25887397},
}

@article{zhang2023compilingplane,
  author            = {Zhang, X and Cui, L and Tso, FP and Jia, W},
  journal           = {IEEE Transactions on Services Computing},
  month             = {Jul},
  number            = {4},
  pages             = {2490--2502},
  publisher         = {Institute of Electrical and Electronics Engineers (IEEE)},
  title             = {Compiling Service Function Chains via Fine-Grained Composition in the Programmable Data Plane},
  url               = {http://dx.doi.org/10.1109/tsc.2023.3242072},
  volume            = {16},
  year              = {2023},
  doi               = {https://doi.org/10.1109/tsc.2023.3242072},
  eissn             = {1939-1374},
  day               = {1},
  publicationstatus = {published},
}

@article{zhang2023idapperilearning,
  author            = {Zhang, X and Cui, L and Tso, FP and Li, Z and Jia, W},
  journal           = {IEEE Transactions on Services Computing},
  month             = {Jul},
  number            = {4},
  pages             = {2532--2544},
  publisher         = {Institute of Electrical and Electronics Engineers (IEEE)},
  title             = {<i>Dapper:</i> Deploying Service Function Chains in the Programmable Data Plane Via Deep Reinforcement Learning},
  url               = {http://dx.doi.org/10.1109/tsc.2023.3237244},
  volume            = {16},
  year              = {2023},
  doi               = {https://doi.org/10.1109/tsc.2023.3237244},
  eissn             = {1939-1374},
  day               = {1},
  publicationstatus = {published},
}

The advent of Network Function Virtualization (NFV) and Service Function Chains (SFCs) unleashes the power of dynamic creation of network services using Virtual Network Functions (VNFs). This is of great interest to network operators since poor service quality and resource wastage can potentially hurt their revenue in the long term. However, the study shows with a set of test-bed experiments that packet loss at certain positions (i.e., different VNFs) in an SFC can cause various degrees of resource wastage and performance degradation because of repeated upstream processing and transmission of retransmitted packets. To overcome this challenge, this study focuses on resource scheduling and deployment of SFCs while considering packet loss positions. This study developed a novel SFC packet dropping cost model and formulated an SFC scheduling problem that aims to minimize overall packet dropping cost as a Mixed-Integer Linear Programming (MILP) and proved that it is NP-hard. In this study, Palosis proposed as an efficient scheme in exploiting the functional characteristics of VNFs and their positions in SFCs for scheduling resources and deployment to optimize packet dropping cost. Extensive experiment results show that Palos can achieve up to 42.73% improvement on packet dropping cost and up to 33.03% reduction on average SFC latency when compared with two other state-of-the-art schemes.

@article{liang2023positionawareplacement,
  author    = {Liang, W and Li, C and Cui, L and Tso, FP},
  journal   = {Digital Communications and Networks},
  month     = {Jan},
  publisher = {Elsevier},
  title     = {Position-aware packet loss optimization on service function chain placement},
  url       = {https://doi.org/10.1016/j.dcan.2023.01.002},
  year      = {2023},
  abstract  = {The advent of Network Function Virtualization (NFV) and Service Function Chains (SFCs) unleashes the power of dynamic creation of network services using Virtual Network Functions (VNFs). This is of great interest to network operators since poor service quality and resource wastage can potentially hurt their revenue in the long term. However, the study shows with a set of test-bed experiments that packet loss at certain positions (i.e., different VNFs) in an SFC can cause various degrees of resource wastage and performance degradation because of repeated upstream processing and transmission of retransmitted packets.
               To overcome this challenge, this study focuses on resource scheduling and deployment of SFCs while considering packet loss positions. This study developed a novel SFC packet dropping cost model and formulated an SFC scheduling problem that aims to minimize overall packet dropping cost as a Mixed-Integer Linear Programming (MILP) and proved that it is NP-hard. In this study, Palosis proposed as an efficient scheme in exploiting the functional characteristics of VNFs and their positions in SFCs for scheduling resources and deployment to optimize packet dropping cost. Extensive experiment results show that Palos can achieve up to 42.73\% improvement on packet dropping cost and up to 33.03\% reduction on average SFC latency when compared with two other state-of-the-art schemes.},
  doi       = {https://doi.org/10.1016/j.dcan.2023.01.002},
  issn      = {2468-5925},
  eissn     = {2352-8648},
  keyword   = {3303 Design},
  keyword   = {4006 Communications engineering},
  keyword   = {4606 Distributed computing and systems software},
  keyword   = {Network function virtualization},
  keyword   = {Resource scheduling},
  keyword   = {SFC deployment},
  keyword   = {Packet loss},
  language  = {English},
  day       = {13},
  version   = {Figshare item version: 1},
}

@inproceedings{chen2023scachecomputing,
  author            = {Chen, C and Nagel, L and Cui, L and Tso, FP},
  booktitle         = {Proceedings of the 6th International Workshop on Edge Systems, Analytics and Networking},
  month             = {May},
  publisher         = {ACM},
  title             = {S-Cache: Function Caching for Serverless Edge Computing},
  url               = {http://dx.doi.org/10.1145/3578354.3592865},
  year              = {2023},
  doi               = {https://doi.org/10.1145/3578354.3592865},
  conference        = {EdgeSys '23: 6th International Workshop on Edge Systems, Analytics and Networking},
  day               = {8},
  publicationstatus = {published},
}

With the ever-growing demand for low-latency network applications, edge computing emerges as a new paradigm that provides computation and storage resources in close proximity to end-users. Many research efforts have resorted to network function virtualization, wherein network applications are provisioned as service function chains at edge clouds. However, due to the traffic dynamics and limited resource capacity at the network edge, how to efficiently embed service chains with latency optimization and resource efficiency remains as a challenging problem. As most existing research efforts largely overlook the bottlenecked resources of VNFs in the VNF scaling, we seek a more realistic approach to provisioning VNF instances across multiple edge clouds. Also, given the limited resources at the edge, it is of significant importance to improve the VNF utilization rate. Specifically, we formulate the VNF scaling problem as an integer linear programming (ILP) problem, aiming to minimize the endto-end latency for service function chains. To solve this problem, we devise a novel bottleneck-aware algorithm that manages the number and deployment of newly created instances. After that, we propose an online algorithm for traffic steering to improve the utilization rates of VNF instances and avoid congestion on hotspot links. The proposed algorithm is shown to provide good performance by trace-driven simulation in real-world topologies.

@inproceedings{chen2022bscaleclouds,
  author       = {Chen, C and Nagel, L and Cui, L and Tso, FP},
  booktitle    = {},
  month        = {Oct},
  organization = {Rhodes, Greece},
  publisher    = {IEEE},
  title        = {B-Scale: Bottleneck-aware VNF scaling and flow routing in edge clouds},
  url          = {https://doi.org/10.1109/ISCC55528.2022.9912772},
  year         = {2022},
  abstract     = {With the ever-growing demand for low-latency network applications, edge computing emerges as a new paradigm that provides computation and storage resources in close proximity to end-users. Many research efforts have resorted to network function virtualization, wherein network applications are provisioned as service function chains at edge clouds. However, due to the traffic dynamics and limited resource capacity at the network edge, how to efficiently embed service chains with latency optimization and resource efficiency remains as a challenging problem. As most existing research efforts largely overlook the bottlenecked resources of VNFs in the VNF scaling, we seek a more realistic approach to provisioning VNF instances across multiple edge clouds. Also, given the limited resources at the edge, it is of significant importance to improve the VNF utilization rate. Specifically, we formulate the VNF scaling problem as an integer linear programming (ILP) problem, aiming to minimize the endto-end latency for service function chains. To solve this problem, we devise a novel bottleneck-aware algorithm that manages the number and deployment of newly created instances. After that, we propose an online algorithm for traffic steering to improve the utilization rates of VNF instances and avoid congestion on hotspot links. The proposed algorithm is shown to provide good performance by trace-driven simulation in real-world topologies.},
  doi          = {https://doi.org/10.1109/ISCC55528.2022.9912772},
  startyear    = {2022},
  startmonth   = {Jun},
  startday     = {30},
  finishyear   = {2022},
  finishmonth  = {Jul},
  finishday    = {3},
  isbn         = {9781665497923},
  eissn        = {2642-7389},
  keyword      = {Cloud computing},
  keyword      = {Service function chaining},
  keyword      = {Integer linear programming},
  keyword      = {Routing},
  keyword      = {010112 Topology},
  keyword      = {Network function virtualization},
  keyword      = {Resource management},
  language     = {English},
  conference   = {27th IEEE Symposium on Computers and Communications (ISCC 2022); 2022 IEEE Symposium on Computers and Communications (ISCC)},
  day          = {19},
  version      = {Figshare item version: 1},
}

Silent packet drops are common in data center networks, and are a major cause of network performance anomalies (NPAs) that have significant impacts on application performance and network management. However, existing solutions using coarse-grained statistics and flow-level telemetry either fail to provide precise location of packet drops or incur large overhead. This paper presents dDrops, a packet-level telemetry based on programmable data plane to detect and retrieve details of silent packet drops immediately when they happen. dDrops can dynamically adapt to varying ratios of silent packet drops for different ports on a switch to improve performance of silent packet drops detection. Moreover, a dynamic memory management scheme is also designed to efficiently use the limited memory on the data plane of switch. dDrops has been implemented on both P4 hardware programmable switches (based on Intel Tofino ASIC) and BMv2. Extensive experiment results show that dDrops is able to detect and locate the silent packet drops within 5 ms (including detailed information of dropped packets), and reduce the memory consumption by up to 50%.

@article{qian2022ddropsplane,
  author    = {Qian, M and Cui, L and Zhang, X and Tso, FP and Deng, Y},
  journal   = {Computer Networks},
  month     = {Jul},
  number    = {2022},
  publisher = {Elsevier},
  title     = {dDrops: Detecting silent packet drops on programmable data plane},
  url       = {https://doi.org/10.1016/j.comnet.2022.109171},
  volume    = {214},
  year      = {2022},
  abstract  = {Silent packet drops are common in data center networks, and are a major cause of network performance anomalies (NPAs) that have significant impacts on application performance and network management. However, existing solutions using coarse-grained statistics and flow-level telemetry either fail to provide precise location of packet drops or incur large overhead. This paper presents dDrops, a packet-level telemetry based on programmable data plane to detect and retrieve details of silent packet drops immediately when they happen. dDrops can dynamically adapt to varying ratios of silent packet drops for different ports on a switch to improve performance of silent packet drops detection. Moreover, a dynamic memory management scheme is also designed to efficiently use the limited memory on the data plane of switch. dDrops has been implemented on both P4 hardware programmable switches (based on Intel Tofino ASIC) and BMv2. Extensive experiment results show that dDrops is able to detect and locate the silent packet drops within 5 ms (including detailed information of dropped packets), and reduce the memory consumption by up to 50\%.},
  doi       = {https://doi.org/10.1016/j.comnet.2022.109171},
  issn      = {1389-1286},
  keyword   = {Silent packet drop},
  keyword   = {Packet-level telemetry},
  keyword   = {Programmable data plane},
  language  = {English},
  day       = {14},
  version   = {Figshare item version: 1},
}

Future networks are expected to support cross-domain, cost-aware and fine-grained services in an efficient and flexible manner. Service Function Chaining (SFC) has been introduced as a promising approach to deliver these services. In the literature, centralized resource orchestration is usually employed to process SFC requests and manage computing and network resources. However, centralized approaches inhibit the scalability and domain autonomy in multi-domain networks. They also neglect location and hardware dependencies of service chains. In this paper, we propose Distributed Federated Service Chaining (DFSC), a framework for orchestrating and maintaining SFC placement in a distributed fashion while sharing only a minimal amount of domain information and control. First, a deployment cost minimization problem is formulated as an Integer Linear Programming (ILP) problem with fine-grained constraints for location and hardware dependencies. We show that this problem is NP-hard. Then, a placement algorithm is devised to use information only on inter-domain paths and border nodes. Our extensive experimental results demonstrate that DFSC efficiently optimizes the deployment cost, supports domain autonomy and enables faster decision-making. The results also show that DFSC finds solutions within a factor 1.15 of the optimal solution on average. Compared to a centralized approach in the literature, DFSC reduces the deployment cost by up to 20% and uses 70% less decision-making time.

@article{chen2022distributednetworks,
  author    = {Chen, C and Nagel, L and Cui, L and Tso, FP},
  journal   = {Computer Networks},
  month     = {May},
  number    = {109044},
  publisher = {Elsevier},
  title     = {Distributed federated service chaining: A scalable and cost-aware approach for multi-domain networks},
  url       = {https://doi.org/10.1016/j.comnet.2022.109044},
  volume    = {212},
  year      = {2022},
  abstract  = {Future networks are expected to support cross-domain, cost-aware and fine-grained services in an efficient and flexible manner. Service Function Chaining (SFC) has been introduced as a promising approach to deliver these services. In the literature, centralized resource orchestration is usually employed to process SFC requests and manage computing and network resources. However, centralized approaches inhibit the scalability and domain autonomy in multi-domain networks. They also neglect location and hardware dependencies of service chains.
               In this paper, we propose Distributed Federated Service Chaining (DFSC), a framework for orchestrating and maintaining SFC placement in a distributed fashion while sharing only a minimal amount of domain information and control. First, a deployment cost minimization problem is formulated as an Integer Linear Programming (ILP) problem with fine-grained constraints for location and hardware dependencies. We show that this problem is NP-hard. Then, a placement algorithm is devised to use information only on inter-domain paths and border nodes. Our extensive experimental results demonstrate that DFSC efficiently optimizes the deployment cost, supports domain autonomy and enables faster decision-making. The results also show that DFSC finds solutions within a factor 1.15 of the optimal solution on average. Compared to a centralized approach in the literature, DFSC reduces the deployment cost by up to 20\% and uses 70\% less decision-making time.},
  doi       = {https://doi.org/10.1016/j.comnet.2022.109044},
  issn      = {1389-1286},
  keyword   = {Multi-domain},
  keyword   = {Distributed orchestration},
  keyword   = {Service function chaining},
  language  = {English},
  day       = {18},
  version   = {Figshare item version: 1},
}

Multi-access Edge Computing (MEC) offers cloud computing capabilities at the edge of the network. Growing demand for low?latency services requires Service Function Chains (SFCs) to be scaled up beyond MEC network to core network. To adapt to network dynamics and provide low-latency services, being able to migrate SFCs when needed is of paramount importance. However, migration of SFCs among edge and core networks such that average latency is optimized as well as considering resource consumption is an intractable challenge because improper migration of Virtual Network Functions (VNFs) results in failure of meeting the requirements of network policies. In this paper, we investigate SFCs in edge-core networks and model the Latency-aware Edge-Core SFCs Migration problem based on open Jackson networks. Two SFC migration algorithms, i.e., Profit-driven Heuristic Search (PHS) and Average Utilization Based (AUB), are proposed to efficiently optimize average latency of all SFCs in edge-core networks. Extensive evaluation results show that PHS optimizes average latency by 19.5%, while AUB can further reduce average latency by up to 36.9% by allowing a marginally higher number of VNF migrations.

@article{liang2022lowlatencynetworks,
  author    = {Liang, W and Cui, L and Tso, FP},
  journal   = {Journal of Systems Architecture},
  month     = {Jan},
  number    = {102405},
  publisher = {Elsevier BV},
  title     = {Low-latency service function chain migration in edge-core networks based on open Jackson networks},
  url       = {https://doi.org/10.1016/j.sysarc.2022.102405},
  volume    = {124},
  year      = {2022},
  abstract  = {Multi-access Edge Computing (MEC) offers cloud computing capabilities at the edge of the network. Growing demand for low?latency services requires Service Function Chains (SFCs) to be scaled up beyond MEC network to core network. To adapt to
               network dynamics and provide low-latency services, being able to migrate SFCs when needed is of paramount importance. However,
               migration of SFCs among edge and core networks such that average latency is optimized as well as considering resource consumption
               is an intractable challenge because improper migration of Virtual Network Functions (VNFs) results in failure of meeting the
               requirements of network policies. In this paper, we investigate SFCs in edge-core networks and model the Latency-aware Edge-Core
               SFCs Migration problem based on open Jackson networks. Two SFC migration algorithms, i.e., Profit-driven Heuristic Search
               (PHS) and Average Utilization Based (AUB), are proposed to efficiently optimize average latency of all SFCs in edge-core networks.
               Extensive evaluation results show that PHS optimizes average latency by 19.5\%, while AUB can further reduce average latency by
               up to 36.9\% by allowing a marginally higher number of VNF migrations.},
  doi       = {https://doi.org/10.1016/j.sysarc.2022.102405},
  issn      = {1383-7621},
  keyword   = {461299 Software engineering not elsewhere classified},
  keyword   = {Computer Hardware \& Architecture},
  keyword   = {Latency},
  keyword   = {Core Network},
  keyword   = {MEC},
  keyword   = {SFC Migration},
  keyword   = {0803 Computer Software},
  language  = {English},
  day       = {21},
  version   = {Figshare item version: 1},
}

The easy exploitation of IoT devices with limited security, compute and processing power has enabled hackers to carry out sophisticated attacks. Many research studies have highlighted the benefits of utilising artificial-intelligence based models in DDoS detection, but emphasis has not been placed on quantitative measurements of compute requirements for Machine Learning and Deep Learning algorithms used for DDoS detection, especially in the inference or detection stage. This research aims to fill the gap by performing quantitative measurement and comparison of various lightweight ML and DL algorithms, as well as design a lightweight collaborative framework capable of DDoS detection close to the source of the attack.

@inproceedings{sofoluwe2022mitigatingedge,
  author            = {Sofoluwe, T and Tso, FP and Phillips, I},
  booktitle         = {Proceedings of the ACM SIGCOMM Internet Measurement Conference, IMC},
  month             = {Oct},
  pages             = {776--777},
  title             = {Mitigating Cyber Threats at the Network Edge},
  year              = {2022},
  abstract          = {The easy exploitation of IoT devices with limited security, compute and processing power has enabled hackers to carry out sophisticated attacks. Many research studies have highlighted the benefits of utilising artificial-intelligence based models in DDoS detection, but emphasis has not been placed on quantitative measurements of compute requirements for Machine Learning and Deep Learning algorithms used for DDoS detection, especially in the inference or detection stage. This research aims to fill the gap by performing quantitative measurement and comparison of various lightweight ML and DL algorithms, as well as design a lightweight collaborative framework capable of DDoS detection close to the source of the attack.},
  doi               = {https://doi.org/10.1145/3517745.3563034},
  isbn              = {9781450392594},
  day               = {25},
  publicationstatus = {published},
}

As a novel UDP-based transport protocol which supports stream multiplexing, QUIC is faster, more lightweight and flexible than TCP. With the prevalence of multi-homed devices such as smartphones with both WiFi and 4G/5G cellular connectivity, Multipath QUIC (MPQUIC) can effectively utilize multiple network interfaces (i.e., multiple paths) to improve transmission efficiency. Current MPQUIC implementation adopts the Lowest-RTT-First (LRF) scheduler which always selects the path with the lowest smoothed RTT among all available paths. However, we show that in networks with heterogeneous paths where network characteristics (e.g., RTT, loss rate) differ considerably, such scheduling scheme leads to unnecessary waiting on fast paths and bufferbloat, degrading overall transmission performance significantly. To use heterogeneous paths efficiently (i.e., to reduce the overall file transfer completion time), this paper proposes a novel scheduling mechanism that assigns data to paths with transfer simulation without causing much additional overhead. Extensive experiment results in Mininet demonstrate that the proposed scheduling mechanism can reduce the transfer completion time by up to 29.6% as compared to existing MPQUIC implementation.

@article{zeng2022optimizingpaths,
  author    = {Zeng, H and Cui, L and Tso, FP and Zhang, Z},
  journal   = {Computer Networks},
  month     = {Jul},
  number    = {2022},
  publisher = {Elsevier},
  title     = {Optimizing multipath QUIC transmission over heterogeneous paths},
  url       = {https://doi.org/10.1016/j.comnet.2022.109198},
  volume    = {215},
  year      = {2022},
  abstract  = {As a novel UDP-based transport protocol which supports stream multiplexing, QUIC is faster, more lightweight and flexible than TCP. With the prevalence of multi-homed devices such as smartphones with both WiFi and 4G/5G cellular connectivity, Multipath QUIC (MPQUIC) can effectively utilize multiple network interfaces (i.e., multiple paths) to improve transmission efficiency. Current MPQUIC implementation adopts the Lowest-RTT-First (LRF) scheduler which always selects the path with the lowest smoothed RTT among all available paths. However, we show that in networks with heterogeneous paths where network characteristics (e.g., RTT, loss rate) differ considerably, such scheduling scheme leads to unnecessary waiting on fast paths and bufferbloat, degrading overall transmission performance significantly. To use heterogeneous paths efficiently (i.e., to reduce the overall file transfer completion time), this paper proposes a novel scheduling mechanism that assigns data to paths with transfer simulation without causing much additional overhead. Extensive experiment results in Mininet demonstrate that the proposed scheduling mechanism can reduce the transfer completion time by up to 29.6\% as compared to existing MPQUIC implementation.},
  doi       = {https://doi.org/10.1016/j.comnet.2022.109198},
  issn      = {1389-1286},
  keyword   = {Transport Protocol},
  keyword   = {Multipath QUIC},
  keyword   = {Scheduling},
  language  = {English},
  day       = {19},
  version   = {Figshare item version: 1},
}

Dynamic service function chains (SFC) are enabled by network function virtualization on general purpose servers. The emergence of programmable data planes (PDP) has offered a new way for the deployment of SFC. However, the implementation of network functions is constrained by resource limitations in PDPs (e.g., compute and memory resource). Moreover, most of existing works do not consider the optimization of state information (e.g., registers), which is essential for stateful network functions. In this paper, we propose pSFC which provides a fine-grained SFC deployment scheme in the PDP to tackle the problem. We first model network functions as control flow graphs (CFG) and the process of deployment as a one big switch (OBS) problem, and then propose an ILP (Integer Linear Programming) model for resource optimization for the OBS problem, which is NP-hard. To solve this problem efficiently, pSFC first composes multiple SFCs for eliminating redundant resources, decomposes the compound CFG based on the resource limitation per stage, and finally maps OBS into the substrate network. We have implemented pSFC in both bmv2 software switch and P4 hardware switch (i.e., Intel Tofino). Evaluation shows that pSFC reduces switch costs 45.7% and average latency 15% while providing the correctness of the process of SFC.

@inproceedings{zhang2022psfcplane,
  author       = {Zhang, X and Cui, L and Tso, FP},
  booktitle    = {},
  month        = {Jul},
  organization = {Taormina, Italy},
  pages        = {484--493},
  publisher    = {IEEE},
  title        = {pSFC: Fine-grained composition of service function chains in the programmable data plane},
  url          = {https://doi.org/10.1109/CCGrid54584.2022.00058},
  year         = {2022},
  abstract     = {Dynamic service function chains (SFC) are enabled by network function virtualization on general purpose servers. The emergence of programmable data planes (PDP) has offered a new way for the deployment of SFC. However, the implementation of network functions is constrained by resource limitations in PDPs (e.g., compute and memory resource). Moreover, most of existing works do not consider the optimization of state information (e.g., registers), which is essential for stateful network functions. In this paper, we propose pSFC which provides a fine-grained SFC deployment scheme in the PDP to tackle the problem. We first model network functions as control flow graphs (CFG) and the process of deployment as a one big switch (OBS) problem, and then propose an ILP (Integer Linear Programming) model for resource optimization for the OBS problem, which is NP-hard. To solve this problem efficiently, pSFC first composes multiple SFCs for eliminating redundant resources, decomposes the compound CFG based on the resource limitation per stage, and finally maps OBS into the substrate network. We have implemented pSFC in both bmv2 software switch and P4 hardware switch (i.e., Intel Tofino). Evaluation shows that pSFC reduces switch costs 45.7\% and average latency 15\% while providing the correctness of the process of SFC.},
  doi          = {https://doi.org/10.1109/CCGrid54584.2022.00058},
  startyear    = {2022},
  startmonth   = {May},
  startday     = {16},
  finishyear   = {2022},
  finishmonth  = {May},
  finishday    = {19},
  isbn         = {9781665499569},
  keyword      = {P4},
  keyword      = {Service function chains},
  keyword      = {Programmable data plane},
  keyword      = {One big switch},
  language     = {English},
  conference   = {2022 22nd IEEE International Symposium on Cluster, Cloud and Internet Computing (CCGrid); 2022 22nd IEEE International Symposium on Cluster, Cloud and Internet Computing (CCGrid)},
  day          = {19},
  version      = {Figshare item version: 1},
}

Future networks are expected to support cross-domain, cost-aware and fine-grained services in an efficient and flexible manner. Service Function Chaining (SFC) has been introduced as a promising approach to deliver these services. In the literature, centralized resource orchestration is usually employed to process SFC requests and manage computing and network resources. However, centralized approaches inhibit the scalability and domain autonomy in multi-domain networks. They also neglect location and hardware dependencies of service chains. In this paper, we propose federated service chaining, a distributed framework which orchestrates and maintains the SFC placement while sharing a minimal amount of domain information and control. We first formulate a deployment cost minimization problem as an Integer Linear Programming (ILP) problem with fine-grained constraints for location and hardware dependencies, which is NPhard. We then devise a Distributed Federated Service Chaining placement approach (DFSC) using inter-domain paths and border nodes information. Our extensive experiments demonstrate that DFSC efficiently optimizes the deployment cost, supports domain autonomy and enables faster decision-making. The results show that DFSC finds solutions within a factor 1.15 of the optimal solution. Compared to a centralized approach in the literature, DFSC reduces the deployment cost by 12% while being one order of magnitude faster.

@inproceedings{chen2021distributedenvironments,
  author       = {Chen, C and Nagel, L and Cui, L and Tso, FP},
  booktitle    = {},
  month        = {Dec},
  organization = {Leicester},
  publisher    = {ACM},
  title        = {Distributed federated service chaining for heterogeneous network environments},
  url          = {https://doi.org/10.1145/3468737.3494091},
  year         = {2021},
  abstract     = {Future networks are expected to support cross-domain, cost-aware
                  and fine-grained services in an efficient and flexible manner. Service Function Chaining (SFC) has been introduced as a promising
                  approach to deliver these services. In the literature, centralized
                  resource orchestration is usually employed to process SFC requests
                  and manage computing and network resources. However, centralized approaches inhibit the scalability and domain autonomy in
                  multi-domain networks. They also neglect location and hardware
                  dependencies of service chains.
                  In this paper, we propose federated service chaining, a distributed
                  framework which orchestrates and maintains the SFC placement
                  while sharing a minimal amount of domain information and control.
                  We first formulate a deployment cost minimization problem as
                  an Integer Linear Programming (ILP) problem with fine-grained
                  constraints for location and hardware dependencies, which is NPhard. We then devise a Distributed Federated Service Chaining
                  placement approach (DFSC) using inter-domain paths and border
                  nodes information. Our extensive experiments demonstrate that
                  DFSC efficiently optimizes the deployment cost, supports domain
                  autonomy and enables faster decision-making. The results show
                  that DFSC finds solutions within a factor 1.15 of the optimal solution.
                  Compared to a centralized approach in the literature, DFSC reduces
                  the deployment cost by 12\% while being one order of magnitude
                  faster.},
  doi          = {https://doi.org/10.1145/3468737.3494091},
  startyear    = {2021},
  startmonth   = {Dec},
  startday     = {6},
  finishyear   = {2021},
  finishmonth  = {Dec},
  finishday    = {9},
  isbn         = {9781450385640},
  keyword      = {Federated Service Chaining},
  keyword      = {Affinity},
  keyword      = {Distributed orchestration},
  keyword      = {Multi-domain},
  language     = {English},
  conference   = {UCC '21: 2021 IEEE/ACM 14th International Conference on Utility and Cloud Computing; UCC '21: Proceedings of the 14th IEEE/ACM International Conference on Utility and Cloud Computing},
  day          = {6},
  version      = {Figshare item version: 1},
}

This paper investigates a dynamic 3D trajectory design of multiple cache-enabled unmanned aerial vehicles (UAVs) in a wireless device-to-device (D2D) caching network with the goal of maximizing the long-term network throughput. By storing popular content at the nearby mobile user devices, D2D caching is an efficient method to improve network throughput and alleviate backhaul burden. With the attractive features of high mobility and flexible deployment, UAVs have recently attracted significant attention as cache-enabled flying base stations. The use of cache-enabled UAVs opens up the possibility of tracking the mobility pattern of the corresponding users and serving them under limited cache storage capacity. However, it is challenging to determine the optimal UAV trajectory due to the dynamic environment with frequently changing network topology and the coexistence of aerial and terrestrial caching nodes. In response, we propose a novel multi-agent reinforcement learning based framework to determine the optimal 3D trajectory of each UAV in a distributed manner without a central coordinator. In the proposed method, multiple UAVs can cooperatively make flight decisions by sharing the gained experiences within a certain proximity to each other. Simulation results reveal that our algorithm outperforms the traditional single- and multi-agent Q-learning algorithms. This work confirms the feasibility and effectiveness of cache-enabled UAVs which serve as an important complement to terrestrial D2D caching nodes.

@article{chen2021multiagentnetworks,
  author    = {Chen, Y-J and Liao, K-M and Ku, M-L and Tso, FP and Chen, G-Y},
  journal   = {IEEE Transactions on Vehicular Technology},
  month     = {Jul},
  number    = {8},
  pages     = {8201--8215},
  publisher = {Institute of Electrical and Electronics Engineers},
  title     = {Multi-agent reinforcement learning based 3D trajectory design in aerial-terrestrial wireless caching networks},
  url       = {https://doi.org/10.1109/TVT.2021.3094273},
  volume    = {70},
  year      = {2021},
  abstract  = {This paper investigates a dynamic 3D trajectory design of multiple cache-enabled unmanned aerial vehicles (UAVs) in a wireless device-to-device (D2D) caching network with the goal of maximizing the long-term network throughput. By storing popular content at the nearby mobile user devices, D2D caching is an efficient method to improve network throughput and alleviate backhaul burden. With the attractive features of high mobility and flexible deployment, UAVs have recently attracted significant attention as cache-enabled flying base stations. The use of cache-enabled UAVs opens up the possibility of tracking the mobility pattern of the corresponding users and serving them under limited cache storage capacity. However, it is challenging to determine the optimal UAV trajectory due to the dynamic environment with frequently changing network topology and the coexistence of aerial and terrestrial caching nodes. In response, we propose a novel multi-agent reinforcement learning based framework to determine the optimal 3D trajectory of each UAV in a distributed manner without a central coordinator. In the proposed method, multiple UAVs can cooperatively make flight decisions by sharing the gained experiences within a certain proximity to each other. Simulation results reveal that our algorithm outperforms the traditional single- and multi-agent Q-learning algorithms. This work confirms the feasibility and effectiveness of cache-enabled UAVs which serve as an important complement to terrestrial D2D caching nodes.},
  doi       = {https://doi.org/10.1109/TVT.2021.3094273},
  issn      = {0018-9545},
  eissn     = {1939-9359},
  keyword   = {unmanned aerial vehicles (UAVs)},
  keyword   = {trajectory design},
  keyword   = {wireless caching},
  keyword   = {multi-agent reinforcement learning},
  language  = {English},
  day       = {2},
  version   = {Figshare item version: 1},
}

Since heavy flows account for a significant fraction of network traffic, being able to predict heavy flows has benefited many network management applications for mitigating link congestion, scheduling of network capacity, exposing network attacks and so on. Existing machine learning based predictors are largely implemented on the control plane of Software Defined Networking (SDN) paradigm. As a result, frequent communication between the control and data planes can cause unnecessary overhead and additional delay in decision making. In this paper, we present pHeavy, a machine learning based scheme for predicting heavy flows directly on the programmable data plane, thus eliminating network overhead and latency to SDN controller. Considering the scarce memory and limited computation capability in the programmable data plane, pHeavy includes a packet processing pipeline which deploys pre-trained decision tree models for in-network prediction. We have implemented pHeavy in both bmv2 software switch and P4 hardware switch (i.e., Barefoot Tofino).Evaluation results demonstrate that pHeavy has achieved 85% and 98% accuracy after receiving the first 5 and 20 packets of a flow respectively, while being able to reduce the size of decision tree by 5.4x on average. More importantly, pHeavy can predict heavy flows at line rate on the P4 hardware switch.

@article{zhang2021pheavyplane,
  author    = {Zhang, X and Cui, L and Tso, FP and Jia, W},
  journal   = {IEEE Transactions on Network and Service Management},
  month     = {Jul},
  number    = {4},
  pages     = {4353--4364},
  publisher = {IEEE},
  title     = {pHeavy: Predicting heavy flows in the programmable data plane},
  url       = {https://doi.org/10.1109/TNSM.2021.3094514},
  volume    = {18},
  year      = {2021},
  abstract  = {Since heavy flows account for a significant fraction
               of network traffic, being able to predict heavy flows has benefited
               many network management applications for mitigating link
               congestion, scheduling of network capacity, exposing network
               attacks and so on. Existing machine learning based predictors
               are largely implemented on the control plane of Software Defined
               Networking (SDN) paradigm. As a result, frequent communication between the control and data planes can cause unnecessary
               overhead and additional delay in decision making.
               In this paper, we present pHeavy, a machine learning based
               scheme for predicting heavy flows directly on the programmable
               data plane, thus eliminating network overhead and latency to
               SDN controller. Considering the scarce memory and limited
               computation capability in the programmable data plane, pHeavy
               includes a packet processing pipeline which deploys pre-trained
               decision tree models for in-network prediction. We have implemented pHeavy in both bmv2 software switch and P4 hardware
               switch (i.e., Barefoot Tofino).Evaluation results demonstrate that
               pHeavy has achieved 85\% and 98\% accuracy after receiving the
               first 5 and 20 packets of a flow respectively, while being able
               to reduce the size of decision tree by 5.4x on average. More
               importantly, pHeavy can predict heavy flows at line rate on the
               P4 hardware switch.},
  doi       = {https://doi.org/10.1109/TNSM.2021.3094514},
  eissn     = {1932-4537},
  keyword   = {460699 Distributed computing and systems software not elsewhere classified},
  keyword   = {Networking \& Telecommunications},
  keyword   = {Electrical and Electronic Engineering},
  keyword   = {Communications Technologies},
  keyword   = {P4},
  keyword   = {Programmable data plane},
  keyword   = {Decision tree},
  keyword   = {Heavy flow},
  keyword   = {0805 Distributed Computing},
  language  = {English},
  day       = {5},
  version   = {Figshare item version: 1},
}

V. Software Defined Networking (SDN), which decouples control plane and data plane, normally stores states on controllers to provide flexible programmability and convenient management. However, recent studies have shown that such configuration may cause frequent and unnecessary interactions between data plane and controllers in some cases. For example, a DDoS detection installed on a controller needs to fetch information from data plane periodically, introducing additional network delay and controller overhead. Thus, stateful data plane is proposed to offload states and operation logics from controller to data plane. Stateful data plane allows switches to perform some operations independently, accelerating packets processing while reducing overhead on both controllers and networks. However, stateful data plane increases the complexity of network devices and imposes many new challenges to the management and schedule of SDN enabled networks. This paper conducts a comprehensive survey on the latest research works and provides insights into stateful data plane. Both stateful data plane platforms and compilers are extensively summarized and analyzed, as well as explicit design of applications based on them. Afterwards, we dwell on the fundamental technologies and research challenges, including implementation considerations of stateful data plane. Finally, we conclude this survey paper with some future works and discuss open research issues.

@article{zhang2020anetworks,
  author            = {Zhang, X and Cui, L and Wei, K and Tso, FP and Ji, Y and Jia, W},
  journal           = {Computer Networks},
  month             = {Oct},
  publisher         = {Elsvier},
  title             = {A survey on stateful data plane in software defined networks},
  year              = {2020},
  abstract          = {© 2020 Elsevier B.V. Software Defined Networking (SDN), which decouples control plane and data plane, normally stores states on controllers to provide flexible programmability and convenient management. However, recent studies have shown that such configuration may cause frequent and unnecessary interactions between data plane and controllers in some cases. For example, a DDoS detection installed on a controller needs to fetch information from data plane periodically, introducing additional network delay and controller overhead. Thus, stateful data plane is proposed to offload states and operation logics from controller to data plane. Stateful data plane allows switches to perform some operations independently, accelerating packets processing while reducing overhead on both controllers and networks. However, stateful data plane increases the complexity of network devices and imposes many new challenges to the management and schedule of SDN enabled networks. This paper conducts a comprehensive survey on the latest research works and provides insights into stateful data plane. Both stateful data plane platforms and compilers are extensively summarized and analyzed, as well as explicit design of applications based on them. Afterwards, we dwell on the fundamental technologies and research challenges, including implementation considerations of stateful data plane. Finally, we conclude this survey paper with some future works and discuss open research issues.},
  doi               = {https://doi.org/10.1016/j.comnet.2020.107597},
  issn              = {1389-1286},
  day               = {10},
  publicationstatus = {published},
}

Unmanned aerial vehicles (UAVs), aka drones, are widely used civil and commercial applications. A promising one is to use the drones as relying nodes to extend the wireless coverage. However, existing solutions only focus on deploying them to predefined locations. After that, they either remain stationary or only move in predefined trajectories throughout the whole deployment. In the open outdoor scenarios such as search and rescue or large music events, etc., users can move and cluster dynamically. As a result, network demand will change constantly over time and hence will require the drones to adapt dynamically. In this paper, we present a proof of concept implementation of an UAV access point (AP) which can dynamically reposition itself depends on the users movement on the ground. Our solution is to continuously keeping track of the received signal strength from the user devices for estimating the distance between users devices and the drone, followed by trilateration to localise them. This process is challenging because our on-site measurements show that the heterogeneity of user devices means that change of their signal strengths reacts very differently to the change of distance to the drone AP. Our initial results demonstrate that our drone is able to effectively localise users and autonomously moving to a position closer to them.

@inproceedings{nunns2020autonomouspoint,
  author       = {Nunns, GJ and Chen, Y-J and Chang, D-K and Liao, K-M and Tso, FP and Cui, L},
  booktitle    = {},
  month        = {Jan},
  organization = {Barcelona, Spain},
  pages        = {278--283},
  publisher    = {IEEE},
  title        = {Autonomous flying WiFi access point},
  url          = {https://doi.org/10.1109/ISCC47284.2019.8969672},
  year         = {2020},
  abstract     = {Unmanned aerial vehicles (UAVs), aka drones, are
                  widely used civil and commercial applications. A promising one is
                  to use the drones as relying nodes to extend the wireless coverage.
                  However, existing solutions only focus on deploying them to
                  predefined locations. After that, they either remain stationary
                  or only move in predefined trajectories throughout the whole
                  deployment. In the open outdoor scenarios such as search and
                  rescue or large music events, etc., users can move and cluster
                  dynamically. As a result, network demand will change constantly
                  over time and hence will require the drones to adapt dynamically.
                  In this paper, we present a proof of concept implementation
                  of an UAV access point (AP) which can dynamically reposition
                  itself depends on the users movement on the ground. Our solution
                  is to continuously keeping track of the received signal strength
                  from the user devices for estimating the distance between users
                  devices and the drone, followed by trilateration to localise them.
                  This process is challenging because our on-site measurements
                  show that the heterogeneity of user devices means that change
                  of their signal strengths reacts very differently to the change of
                  distance to the drone AP. Our initial results demonstrate that
                  our drone is able to effectively localise users and autonomously
                  moving to a position closer to them.},
  doi          = {https://doi.org/10.1109/ISCC47284.2019.8969672},
  startyear    = {2019},
  startmonth   = {Jun},
  startday     = {29},
  finishyear   = {2019},
  finishmonth  = {Jul},
  finishday    = {3},
  isbn         = {9781728129990},
  eissn        = {2642-7389},
  keyword      = {469999 Other information and computing sciences not elsewhere classified},
  keyword      = {UAV access point,},
  keyword      = {Flying access point},
  keyword      = {Drone based user localisation},
  keyword      = {089999 Information and Computing Sciences not elsewhere classified},
  language     = {English},
  conference   = {IEEE Symposium on Computers and Communications (ISCC)},
  day          = {27},
  version      = {Figshare item version: 1},
}

Emerging edge computing has seen latency- sensitive services moving rapidly from cloud to the edge to take advantage of its close vicinity to end users, while cloud is retained for carrying out latency-insensitive and computation-intensive tasks. When more edge computing service providers come to the market, the network will become increasingly more fragmented because of their proprietary services and policies deployed in the network. This means that the Internet can become more cumbersome and riskier as there will be more tiers and potential vulnerabilities that could be exploited. To tackle this issue, we envisage a federated service chaining paradigm in which operators can share and put service functions in other participants' networks so as to improve resource utilization, collaboratively mitigate cyber threats, and enable service innovations. In this position article, building on our past experience in enabling federated cloud infrastructure and heterogeneous service chaining, we present a Federated Service Chaining architecture followed by discussions of its key components. Several key research challenges are described for the successful realization of such architecture. We hope this article can open a discussion in the research community and generate enough research interest to significantly advance this field.

@article{cui2020federatedchallenges,
  author            = {Cui, L and Tso, FP and Jia, W},
  journal           = {IEEE Communications Magazine},
  month             = {Mar},
  number            = {3},
  pages             = {47--53},
  publisher         = {IEEE},
  title             = {Federated Service Chaining: Architecture and Challenges},
  volume            = {58},
  year              = {2020},
  abstract          = {© 2020 IEEE. Emerging edge computing has seen latency- sensitive services moving rapidly from cloud to the edge to take advantage of its close vicinity to end users, while cloud is retained for carrying out latency-insensitive and computation-intensive tasks. When more edge computing service providers come to the market, the network will become increasingly more fragmented because of their proprietary services and policies deployed in the network. This means that the Internet can become more cumbersome and riskier as there will be more tiers and potential vulnerabilities that could be exploited. To tackle this issue, we envisage a federated service chaining paradigm in which operators can share and put service functions in other participants' networks so as to improve resource utilization, collaboratively mitigate cyber threats, and enable service innovations. In this position article, building on our past experience in enabling federated cloud infrastructure and heterogeneous service chaining, we present a Federated Service Chaining architecture followed by discussions of its key components. Several key research challenges are described for the successful realization of such architecture. We hope this article can open a discussion in the research community and generate enough research interest to significantly advance this field.},
  doi               = {https://doi.org/10.1109/MCOM.001.1900627},
  issn              = {0163-6804},
  eissn             = {1558-1896},
  day               = {1},
  publicationstatus = {published},
}

The Cloud has become integral to most Internet-based applications and user gadgets. This article provides a brief history of the Cloud and presents a researcher’s view of the prospects for innovating at the infrastructure, middleware, and application and delivery levels of the already crowded Cloud computing stack.

@article{varghese2019cloudfuturology,
  author    = {Varghese, B and Leitner, P and Ray, S and Chard, K and Barker, A and Elkhatib, Y and Herry, H and Hong, C-H and Singer, J and Tso, FP and Yoneki, E and Zhani, M-F},
  journal   = {IEEE Computer},
  month     = {Aug},
  number    = {9},
  pages     = {68--77},
  publisher = {IEEE},
  title     = {Cloud futurology},
  url       = {https://ieeexplore.ieee.org/document/8812159},
  volume    = {52},
  year      = {2019},
  abstract  = {The Cloud has become integral to most Internet-based applications and user gadgets. This
               article provides a brief history of the Cloud and presents a researcher’s view of the prospects
               for innovating at the infrastructure, middleware, and application and delivery levels of the
               already crowded Cloud computing stack.},
  doi       = {https://doi.org/10.1109/MC.2019.2895307},
  issn      = {0018-9162},
  keyword   = {469999 Other information and computing sciences not elsewhere classified},
  keyword   = {089999 Information and Computing Sciences not elsewhere classified},
  language  = {English},
  day       = {23},
  version   = {Figshare item version: 2},
}

In recent decades, many TCP Congestion Control (CC) protocols have been proposed to improve the performance and reliability of TCP in various network scenarios. However, CC protocols are usually closely coupled with network conditions such as latency and packet loss. Considering that networks with different properties are common, e.g., wired/wireless LAN and Long Fat Networks (LFNs), investigating both performance and behaviors of CC protocols under varied network scenarios becomes crucial for both network management and development. In this paper, we conduct a comprehensive measurement study on the goodput, RTT, retransmission, friendliness, fairness, convergence time and stability of most widely-used CC protocols over wired LAN/WAN and wireless LAN (both 2.4GHz and 5GHz Wi-Fi). We also conduct comparative studies with respect to transmission cost, congested reverse path and bottleneck queue size in network simulator. Based on our analysis, we reveal several interesting and original observations. We found that the goodput of BBR is at least 22.5% lower than other CC protocols in wireless LAN due to insufficient pacing rate, even though it can always fully utilize the bottleneck bandwidth with low RTT in wired networks. We also observed that the total on-wire data volume of BBR is higher than CUBIC (e.g., 2.37% higher when RTT = 100ms and loss rate = 0.01%). In addition, BBR can fully utilize the bottleneck bandwidth in most queue sizes (≥ 20packets). Surprisingly, we noticed that as the default CC protocol in most modern operating systems, CUBIC is too aggressive and unfriendly in both LAN and wireless LAN, greatly suppressing the goodput of other competing CC protocols. More specifically for CUBIC in wireless LAN, it generates 129% more retransmissions than other CC protocols. Nevertheless, we have also seen that, in scenario with heavily-congested reverse path, CUBIC can provide full utilization on bottleneck bandwidth. Lastly, we also observed that BBR converges very quickly in all evaluated scenarios, while other CC protocols present varied results, e.g., Westwood+ and Veno converge faster in 5GHz Wi-Fi networks than 2.4GHz networks.

@article{lin2019extensivescenarios,
  author    = {Lin, J and Cui, L and Zhang, Y and Tso, FP and Guan, Q},
  journal   = {Computer Networks},
  month     = {Aug},
  number    = {106872},
  publisher = {Elsevier},
  title     = {Extensive evaluation on the performance and behaviour of TCP congestion control protocols under varied network scenarios},
  volume    = {163},
  year      = {2019},
  abstract  = {In recent decades, many TCP Congestion Control (CC) protocols have been proposed to improve the performance and reliability of TCP in various network scenarios. However, CC protocols are usually closely coupled with network conditions such as latency and packet loss. Considering that networks with different properties are common, e.g., wired/wireless LAN and Long Fat Networks (LFNs), investigating both performance and behaviors of CC protocols under varied network scenarios becomes crucial for both network management and development. In this paper, we conduct a comprehensive measurement study on the goodput, RTT, retransmission, friendliness, fairness, convergence time and stability of most widely-used CC protocols over wired LAN/WAN and wireless LAN (both 2.4GHz and 5GHz Wi-Fi). We also conduct comparative studies with respect to transmission cost, congested reverse path and bottleneck queue size in network simulator. Based on our analysis, we reveal several interesting and original observations. We found that the goodput of BBR is at least 22.5\% lower than other CC protocols in wireless LAN due to insufficient pacing rate, even though it can always fully utilize the bottleneck bandwidth with low RTT in wired networks. We also observed that the total on-wire data volume of BBR is higher than CUBIC (e.g., 2.37\% higher when RTT = 100ms and loss rate = 0.01\%). In addition, BBR can fully utilize the bottleneck bandwidth in most queue sizes (≥ 20packets). Surprisingly, we noticed that as the default CC protocol in most modern operating systems, CUBIC is too aggressive and unfriendly in both LAN and wireless LAN, greatly suppressing the goodput of other competing CC protocols. More specifically for CUBIC in wireless LAN, it generates 129\% more retransmissions than other CC protocols. Nevertheless, we have also seen that, in scenario with heavily-congested reverse path, CUBIC can provide full utilization on bottleneck bandwidth. Lastly, we also observed that BBR converges very quickly in all evaluated scenarios, while other CC protocols present varied results, e.g., Westwood+ and Veno converge faster in 5GHz Wi-Fi networks than 2.4GHz networks.},
  doi       = {https://doi.org/10.1016/j.comnet.2019.106872},
  issn      = {1389-1286},
  keyword   = {TCP Congestion control protocol},
  keyword   = {Performance evaluation},
  language  = {English},
  day       = {5},
  version   = {Figshare item version: 1},
}

This paper investigates the problem of cache node placement and selection with the coexistence of unmanned aerial vehicles (UAVs) cache and device- to-device (D2D) cache in mobile networks. In recent years, caching popular content in UAV base stations has received growing interests as a promising solution to improve communication performances. With the agility and mobility features, the dynamic movement of cache-enabled UAV should be further designed to increase the cache-aided throughput. Different from the conventional caching approaches assuming ground users remain static, we consider the dynamic movement design of UAV to maximize the cache- aided throughput taking into account the movement of ground users. As the formulated optimization problem is NP-hard, we propose a mobility-aware probabilistic caching algorithm in which K-means clustering is utilized to obtain the partition of ground users. Simulation results show that the proposed algorithm notably outperforms the pure D2D cache scheme (without UAV caching) in different cases.

@inproceedings{chen2019mobilityawarenetworks,
  author            = {Chen, YJ and Liao, KM and Ku, ML and Tso, FP},
  booktitle         = {2019 IEEE Global Communications Conference, GLOBECOM 2019 - Proceedings},
  month             = {Dec},
  organization      = {Waikoloa, HI, USA},
  title             = {Mobility-aware probabilistic caching in UAV-assisted wireless D2D networks},
  year              = {2019},
  abstract          = {© 2019 IEEE. This paper investigates the problem of cache node placement and selection with the coexistence of unmanned aerial vehicles (UAVs) cache and device- to-device (D2D) cache in mobile networks. In recent years, caching popular content in UAV base stations has received growing interests as a promising solution to improve communication performances. With the agility and mobility features, the dynamic movement of cache-enabled UAV should be further designed to increase the cache-aided throughput. Different from the conventional caching approaches assuming ground users remain static, we consider the dynamic movement design of UAV to maximize the cache- aided throughput taking into account the movement of ground users. As the formulated optimization problem is NP-hard, we propose a mobility-aware probabilistic caching algorithm in which K-means clustering is utilized to obtain the partition of ground users. Simulation results show that the proposed algorithm notably outperforms the pure D2D cache scheme (without UAV caching) in different cases.},
  doi               = {https://doi.org/10.1109/GLOBECOM38437.2019.9013432},
  startyear         = {2019},
  startmonth        = {Dec},
  startday          = {9},
  finishyear        = {2019},
  finishmonth       = {Dec},
  finishday         = {13},
  isbn              = {9781728109626},
  conference        = {2019 IEEE Global Communications Conference},
  day               = {1},
  publicationstatus = {published},
}

TCP congestion control is a vital component for the latency of Web services. In practice, a single congestion control mechanism is often used to handle all TCP connections on a Web server, e.g., Cubic for Linux by default. Considering complex and ever-changing networking environment, the default congestion control may not always be the most suitable one. Adjusting congestion control to meet different networking scenarios usually requires modification of TCP stacks on a server. This is difficult, if not impossible, due to various operating system and application configurations on production servers. In this paper, we propose Mystique, a light-weight, flexible, and dynamic congestion control switching scheme that allows network or server administrators to deploy any congestion control schemes transparently without modifying existing TCP stacks on servers. We have implemented Mystique in Open vSwitch (OVS) and conducted extensive testbed experiments in both public and private cloud environments. Experiment results have demonstrated that Mystique is able to effectively adapt to varying network conditions, and can always employ the most suitable congestion control for each TCP connection. More specifically, Mystique can significantly reduce latency by 18.13% on average when compared with individual congestion controls.

@article{zhang2019mystiquescheme,
  author            = {Zhang, Y and Cui, L and Tso, FP and Guan, Q and Jia, W and Zhou, J},
  journal           = {IEEE Transactions on Network and Service Management},
  month             = {Aug},
  number            = {4},
  pages             = {1869--1883},
  publisher         = {IEEE},
  title             = {Mystique: a fine-grained and transparent congestion control enforcement scheme},
  volume            = {16},
  year              = {2019},
  abstract          = {TCP congestion control is a vital component for the latency of Web services. In practice, a single congestion control mechanism is often used to handle all TCP connections on a Web server, e.g., Cubic for Linux by default. Considering complex and ever-changing networking environment, the default congestion control may not always be the most suitable one. Adjusting congestion control to meet different networking scenarios usually requires modification of TCP stacks on a server. This is difficult, if not impossible, due to various operating system and application configurations on production servers. In this paper, we propose Mystique, a light-weight, flexible, and dynamic congestion control switching scheme that allows network or server administrators to deploy any congestion control schemes transparently without modifying existing TCP stacks on servers. We have implemented Mystique in Open vSwitch (OVS) and conducted extensive testbed experiments in both public and private cloud environments. Experiment results have demonstrated that Mystique is able to effectively adapt to varying network conditions, and can always employ the most suitable congestion control for each TCP connection. More specifically, Mystique can significantly reduce latency by 18.13\% on average when compared with individual congestion controls.},
  doi               = {https://doi.org/10.1109/TNSM.2019.2935829},
  eissn             = {1932-4537},
  keyword           = {0805 Distributed Computing},
  keyword           = {Web service},
  keyword           = {TCP},
  keyword           = {Congestion control},
  keyword           = {Transfer completion time},
  language          = {English},
  day               = {16},
  publicationstatus = {published},
}

@article{basford2019performanceclusters,
  author            = {Basford, PJ and Johnston, SJ and Perkins, CS and Garnock-Jones, T and Tso, FP and Pezaros, D and Mullins, RD and Yoneki, E and Singer, J and Cox, SJ},
  journal           = {Future Generation Computer Systems},
  month             = {Jul},
  pages             = {278--291},
  publisher         = {Elsevier BV},
  title             = {Performance analysis of single board computer clusters},
  volume            = {102},
  year              = {2019},
  doi               = {https://doi.org/10.1016/j.future.2019.07.040},
  issn              = {0167-739X},
  language          = {en},
  day               = {22},
  publicationstatus = {published},
}

It's said that the pleasure is in the giving, not the receiving. This belief is validated by how humans interact with their family, friends and society as well as their gardens, homes, and pets. Yet for ubiquitous devices, this dynamic is reversed with devices as the donors and owners as the recipients. This paper explores an alternative paradigm where these devices are elevated, becoming members of Data Hungry Homes, allowing us to build relationships with them using the principles that we apply to family, pets or houseplants. These devices are developed to fit into a new concept of the home, can symbiotically interact with us and possess needs and traits that yield unexpected positive or negative outcomes from interacting with them. Such relationships could enrich our lives through our endeavours to “feed” our Data Hungry Homes, possibly leading us to explore new avenues and interactions outside and inside the home.

@inproceedings{leesmith2019thehome,
  author       = {Lee-Smith, M and Ross, T and Maguire, M and Tso, FP and Morley, J and Cavazzi, S},
  booktitle    = {},
  editor       = {Spence, JEFSMMPSRJ},
  month        = {Nov},
  organization = {Nottingham, UK},
  publisher    = {ACM},
  title        = {The data hungry home},
  url          = {https://doi.org/10.1145/3363384.3363390},
  year         = {2019},
  abstract     = {It's said that the pleasure is in the giving, not the receiving. This belief is validated by how humans interact with their family, friends and society as well as their gardens, homes, and pets. Yet for ubiquitous devices, this dynamic is reversed with devices as the donors and owners as the recipients. This paper explores an alternative paradigm where these devices are elevated, becoming members of Data Hungry Homes, allowing us to build relationships with them using the principles that we apply to family, pets or houseplants. These devices are developed to fit into a new concept of the home, can symbiotically interact with us and possess needs and traits that yield unexpected positive or negative outcomes from interacting with them. Such relationships could enrich our lives through our endeavours to “feed” our Data Hungry Homes, possibly leading us to explore new avenues and interactions outside and inside the home.},
  doi          = {https://doi.org/10.1145/3363384.3363390},
  startyear    = {2019},
  startmonth   = {Nov},
  startday     = {19},
  finishyear   = {2019},
  finishmonth  = {Nov},
  finishday    = {20},
  isbn         = {9781450372039},
  keyword      = {Human computer interaction},
  keyword      = {Human data interaction},
  keyword      = {Smart homes},
  keyword      = {Internet of Things},
  keyword      = {Interaction design theory},
  language     = {English},
  conference   = {Halfway to the Future; Proceedings of the Halfway to the Future Symposium 2019 (HTTF 2019)},
  day          = {19},
  version      = {Figshare item version: 1},
}

What is it to navigate or to be navigated? How, and through what, is information communicated to us? Do our interactions with space need to be limited to when we are moving through it? This paper describes a collection of design concepts generated as part of the initial stages of a research project that combines a critical design mindset and research through design process to explore these types of questions. The project seeks to problematise and diversify the discussion and understanding around pedestrian navigation, wearable technology, crowdsourcing and human data interaction. The goal is to develop one of the concepts using research through design as part of PhD research studies, leading to possible future applications.

@inproceedings{leesmith2019whatconcepts,
  author       = {Lee-Smith, M and Ross, T and Maguire, M and Tso, FP and Morley, J and Cavazzi, S},
  booktitle    = {},
  month        = {Jan},
  note         = {© ACM 2019. This is the author's version of the work. It is posted here for your personal use. Not for redistribution. The definitive Version of Record was published in Designing Interactive Systems (DIS) 2019, San Diego, June 23-28, 2019, https://doi.org/10.1145/3301019.3323889},
  organization = {San Diego},
  publisher    = {© ACM},
  title        = {What can we expect from navigating? Exploring navigation, wearables and data through critical design concepts},
  url          = {https://doi.org/10.1145/3301019.3323889},
  year         = {2019},
  abstract     = {What is it to navigate or to be navigated? How, and
                  through what, is information communicated to us? Do
                  our interactions with space need to be limited to when
                  we are moving through it? This paper describes a
                  collection of design concepts generated as part of the
                  initial stages of a research project that combines a
                  critical design mindset and research through design
                  process to explore these types of questions. The project
                  seeks to problematise and diversify the discussion and
                  understanding around pedestrian navigation, wearable
                  technology, crowdsourcing and human data interaction.
                  The goal is to develop one of the concepts using
                  research through design as part of PhD research
                  studies, leading to possible future applications.},
  doi          = {https://doi.org/10.1145/3301019.3323889},
  isbn         = {9781450362702},
  keyword      = {469999 Other information and computing sciences not elsewhere classified},
  keyword      = {Critical design},
  keyword      = {Pedestrian navigation},
  keyword      = {Wearable technology},
  keyword      = {Human data interaction},
  keyword      = {Crowdsourcing},
  keyword      = {089999 Information and Computing Sciences not elsewhere classified},
  language     = {English},
  conference   = {Designing Interactive Systems (DIS) 2019},
  day          = {1},
  version      = {Figshare item version: 1},
}

Copyright held by the owner/author(s). Publication rights licensed to the Association for Computing Machinery. In practice, a single TCP congestion control is often used to handle all TCP connections on a Web server, e.g., Cubic for Linux by default. Considering complex and ever-changing networking environment, the default congestion control algorithm may not always be the most suitable one. Adjusting congestion control usually to meet different networking scenarios requires modification of servers' TCP stacks. This is difficult, if not impossible, due to various operating systems and different configurations on the servers. In this paper, we propose Mystique, a light-weight and flexible scheme that allows administrators (or operators) to deploy any congestion control schemes transparently without changing existing TCP stacks on servers. We have implemented Mystique in Open vSwitch (OVS) and conducted extensive test-bed experiments in public cloud environments. We have extensively evaluated Mystique and the results have demonstrated that it is able to effectively adapt to varying network conditions, and can always employ the most suitable congestion control for each TCP connection. Mystique can significantly reduce latency by up to 37.8% in comparison with other congestion controls.

@inproceedings{zhang2018ascheme,
  author            = {Zhang, Y and Guan, Q and Cui, L and Jia, W and Tso, FP and Zhou, J},
  booktitle         = {ANRW 2018 - Proceedings of the 2018 Applied Networking Research Workshop},
  month             = {Jul},
  organization      = {Montreal},
  pages             = {26--32},
  title             = {A fine-grained and transparent congestion control enforcement scheme},
  year              = {2018},
  abstract          = {© 2018 Copyright held by the owner/author(s). Publication rights licensed to the Association for Computing Machinery. In practice, a single TCP congestion control is often used to handle all TCP connections on a Web server, e.g., Cubic for Linux by default. Considering complex and ever-changing networking environment, the default congestion control algorithm may not always be the most suitable one. Adjusting congestion control usually to meet different networking scenarios requires modification of servers' TCP stacks. This is difficult, if not impossible, due to various operating systems and different configurations on the servers. In this paper, we propose Mystique, a light-weight and flexible scheme that allows administrators (or operators) to deploy any congestion control schemes transparently without changing existing TCP stacks on servers. We have implemented Mystique in Open vSwitch (OVS) and conducted extensive test-bed experiments in public cloud environments. We have extensively evaluated Mystique and the results have demonstrated that it is able to effectively adapt to varying network conditions, and can always employ the most suitable congestion control for each TCP connection. Mystique can significantly reduce latency by up to 37.8\% in comparison with other congestion controls.},
  doi               = {https://doi.org/10.1145/3232755.3232778},
  startyear         = {2018},
  startmonth        = {Jul},
  startday          = {16},
  finishyear        = {2018},
  finishmonth       = {Jul},
  finishday         = {16},
  isbn              = {9781450355858},
  conference        = {Applied Networking Research Workshop 2018},
  day               = {16},
  publicationstatus = {published},
}

Current commodity Single Board Computers (SBCs) are sufficiently powerful to run mainstream operating systems and workloads. Many of these boards may be linked together, to create small, low-cost clusters that replicate some features of large data center clusters. The Raspberry Pi Foundation produces a series of SBCs with a price/performance ratio that makes SBC clusters viable, perhaps even expendable. These clusters are an enabler for Edge/Fog Compute, where processing is pushed out towards data sources, reducing bandwidth requirements and decentralizing the architecture. In this paper we investigate use cases driving the growth of SBC clusters, we examine the trends in future hardware developments, and discuss the potential of SBC clusters as a disruptive technology. Compared to traditional clusters, SBC clusters have a reduced footprint, are low-cost, and have low power requirements. This enables different models of deployment—particularly outside traditional data center environments. We discuss the applicability of existing software and management infrastructure to support exotic deployment scenarios and anticipate the next generation of SBC. We conclude that the SBC cluster is a new and distinct computational deployment paradigm, which is applicable to a wider range of scenarios than current clusters. It facilitates Internet of Things and Smart City systems and is potentially a game changer in pushing application logic out towards the network edge.

@article{johnston2018commodityapplications,
  author            = {Johnston, SJ and Basford, PJ and Perkins, CS and Herry, H and Tso, FP and Pezaros, D and Mullins, RD and Yoneki, E and Cox, SJ and Singer, J},
  journal           = {Future Generation Computer Systems},
  month             = {Dec},
  pages             = {201--212},
  publisher         = {ELsivier},
  title             = {Commodity single board computer clusters and their applications},
  volume            = {89},
  year              = {2018},
  abstract          = {© 2018 Current commodity Single Board Computers (SBCs) are sufficiently powerful to run mainstream operating systems and workloads. Many of these boards may be linked together, to create small, low-cost clusters that replicate some features of large data center clusters. The Raspberry Pi Foundation produces a series of SBCs with a price/performance ratio that makes SBC clusters viable, perhaps even expendable. These clusters are an enabler for Edge/Fog Compute, where processing is pushed out towards data sources, reducing bandwidth requirements and decentralizing the architecture. In this paper we investigate use cases driving the growth of SBC clusters, we examine the trends in future hardware developments, and discuss the potential of SBC clusters as a disruptive technology. Compared to traditional clusters, SBC clusters have a reduced footprint, are low-cost, and have low power requirements. This enables different models of deployment—particularly outside traditional data center environments. We discuss the applicability of existing software and management infrastructure to support exotic deployment scenarios and anticipate the next generation of SBC. We conclude that the SBC cluster is a new and distinct computational deployment paradigm, which is applicable to a wider range of scenarios than current clusters. It facilitates Internet of Things and Smart City systems and is potentially a game changer in pushing application logic out towards the network edge.},
  doi               = {https://doi.org/10.1016/j.future.2018.06.048},
  issn              = {0167-739X},
  day               = {1},
  publicationstatus = {online-published},
}

Network Function Virtualisation (NFV) enables rapid deployment of new services in networks on an on-demand basis using general purpose servers. Multiple virtual network functions (VNFs) can be dynamically chained in an ordered sequence for the delivery of end-to-end services. Nevertheless, network latency caused by the sequential order of packet processing on every VNF can hurt the performance of latency-sensitive applications. To reduce such network latency, existing solutions only consider the maximum capacity of individual virtual network functions (VNFs) and do not take into account the fact that performance of VNFs, as with any software applications, is bottlenecked by either CPU or I/O peripheral capacity of the server they run on and their underneath implementation such as singleor multi-threaded.By exploiting this knowledge, we can better determine the number of required VNF instances and distribute the network traffic among them for any given VNF chain. In this paper, we formulate the VNF Scaling and Traffic Distribution problem and prove that it is NP-hard. We then present the design and implementation of Natif, an efficient VNF-Aware VNF insTantIation and traFfic distribution scheme. Through our OpenStack-based testbed evaluations, we demonstrate that Natif can significantly improve the network latency by 188% on average as compared to other approaches. As a chain composition scheme, Natif can effectively work with any VNF chaining algorithms.

@inproceedings{hajji2018dynamiclatency,
  author            = {Hajji, W and Genez, TAL and Tso, FP and Cui, L and Phillips, I},
  booktitle         = {Proceedings - IEEE Symposium on Computers and Communications},
  month             = {Nov},
  organization      = {Natal, Brazil},
  pages             = {316--321},
  title             = {Dynamic Network Function Chain Composition for Mitigating Network Latency},
  volume            = {2018-June},
  year              = {2018},
  abstract          = {© 2018 IEEE. Network Function Virtualisation (NFV) enables rapid deployment of new services in networks on an on-demand basis using general purpose servers. Multiple virtual network functions (VNFs) can be dynamically chained in an ordered sequence for the delivery of end-to-end services. Nevertheless, network latency caused by the sequential order of packet processing on every VNF can hurt the performance of latency-sensitive applications. To reduce such network latency, existing solutions only consider the maximum capacity of individual virtual network functions (VNFs) and do not take into account the fact that performance of VNFs, as with any software applications, is bottlenecked by either CPU or I/O peripheral capacity of the server they run on and their underneath implementation such as singleor multi-threaded.By exploiting this knowledge, we can better determine the number of required VNF instances and distribute the network traffic among them for any given VNF chain. In this paper, we formulate the VNF Scaling and Traffic Distribution problem and prove that it is NP-hard. We then present the design and implementation of Natif, an efficient VNF-Aware VNF insTantIation and traFfic distribution scheme. Through our OpenStack-based testbed evaluations, we demonstrate that Natif can significantly improve the network latency by 188\% on average as compared to other approaches. As a chain composition scheme, Natif can effectively work with any VNF chaining algorithms.},
  doi               = {https://doi.org/10.1109/ISCC.2018.8538646},
  startyear         = {2018},
  startmonth        = {Jun},
  startday          = {25},
  finishyear        = {2018},
  finishmonth       = {Jun},
  finishday         = {28},
  isbn              = {9781538669501},
  issn              = {1530-1346},
  conference        = {018 IEEE Symposium on Computers and Communications (ISCC)},
  day               = {15},
  publicationstatus = {published},
}

@article{cui2018enablingchaining,
  author            = {Cui, L and Tso, F and Guo, S and Jia, W and Wei, K and Wei, Z},
  journal           = {IEEE Transactions on Parallel and Distributed Systems},
  month             = {Sep},
  publisher         = {IEEE},
  title             = {Enabling Heterogeneous Network Function Chaining},
  year              = {2018},
  doi               = {https://doi.org/10.1109/TPDS.2018.2871845},
  day               = {24},
  publicationstatus = {published},
}

V. Data center operators deploy a variety of both physical and virtual network functions boxes (NFBs) to take advantages of inherent efficiency offered by physical NFBs with the agility and flexibility of virtual ones. However, such heterogeneity faces great challenges in correct, efficient and dynamic network policy implementation because, firstly, existing schemes are limited to exclusively physical or virtual NFBs and not a mix, and secondly, NFBs can co-exist at various locations in the network as a result of emerging technologies such as Software Defined Networking (SDN) and Network Function Virtualization (NFV). In this paper, we propose a Heterogeneous netwOrk pOlicy enforCement scheme (HOOC) to overcome these challenges. We first formulate and model HOOC, which is shown be to NP-Hard by reducing from the Multiple Knapsack Problem (MKP). We then propose an efficient online algorithm that can achieve optimal latency-wise NF service chaining amongst heterogenous NFBs. In addition, we also provide a greedy algorithm when operators prefer smaller run-time than optimality. Our simulation results show that HOOC is efficient and scalable whilst testbed implementation demonstrates that HOOC can be easily deployed in the data center environments.

@article{cui2018enforcingenvironment,
  author            = {Cui, L and Tso, FP and Jia, W},
  journal           = {Computer Networks},
  month             = {Jun},
  pages             = {108--118},
  publisher         = {ELSEVIER},
  title             = {Enforcing network policy in heterogeneous network function box environment},
  volume            = {138},
  year              = {2018},
  abstract          = {© 2018 Elsevier B.V. Data center operators deploy a variety of both physical and virtual network functions boxes (NFBs) to take advantages of inherent efficiency offered by physical NFBs with the agility and flexibility of virtual ones. However, such heterogeneity faces great challenges in correct, efficient and dynamic network policy implementation because, firstly, existing schemes are limited to exclusively physical or virtual NFBs and not a mix, and secondly, NFBs can co-exist at various locations in the network as a result of emerging technologies such as Software Defined Networking (SDN) and Network Function Virtualization (NFV). In this paper, we propose a Heterogeneous netwOrk pOlicy enforCement scheme (HOOC) to overcome these challenges. We first formulate and model HOOC, which is shown be to NP-Hard by reducing from the Multiple Knapsack Problem (MKP). We then propose an efficient online algorithm that can achieve optimal latency-wise NF service chaining amongst heterogenous NFBs. In addition, we also provide a greedy algorithm when operators prefer smaller run-time than optimality. Our simulation results show that HOOC is efficient and scalable whilst testbed implementation demonstrates that HOOC can be easily deployed in the data center environments.},
  doi               = {https://doi.org/10.1016/j.comnet.2018.03.029},
  issn              = {1389-1286},
  day               = {19},
  publicationstatus = {online-published},
}

Network Functions (NFs) are commonly used to provide key security and performance guarantees in networks. However, this is greatly challenged by IP fragments. Since only the first fragment will keep complete layer-4 (L4) header during fragmentation, a firewall that is configured to let this packet through may drop all subsequent fragments due to missing L4 headers, causing policy violations. Also, the incomplete payload in fragments could reduce processing accuracy of NFs (e.g., intrusion detection system) To address these challenges, we present FragGuide, a transparent framework to assist NFs handling fragments correctly efficiently without requiring any modification to NFs. Our experiment results show that FragGuide can achieve 100% accuracy and zero invalid fragments for processing fragments in firewalls, and improve detection rate of IDS by 15.63% on average, while guaranteeing lower processing latency.

@inproceedings{zhang2018fragguidepackets,
  author            = {Zhang, Y and Cui, L and Tso, FP, F},
  booktitle         = {Proceedings of the 2018 IEEE 7th International Conference on Cloud Networking, CloudNet 2018},
  month             = {Nov},
  organization      = {Tokyo, Japan},
  title             = {FragGuide: Enforcing Network Policies on Fragmented Packets},
  year              = {2018},
  abstract          = {© 2018 IEEE. Network Functions (NFs) are commonly used to provide key security and performance guarantees in networks. However, this is greatly challenged by IP fragments. Since only the first fragment will keep complete layer-4 (L4) header during fragmentation, a firewall that is configured to let this packet through may drop all subsequent fragments due to missing L4 headers, causing policy violations. Also, the incomplete payload in fragments could reduce processing accuracy of NFs (e.g., intrusion detection system) To address these challenges, we present FragGuide, a transparent framework to assist NFs handling fragments correctly efficiently without requiring any modification to NFs. Our experiment results show that FragGuide can achieve 100\% accuracy and zero invalid fragments for processing fragments in firewalls, and improve detection rate of IDS by 15.63\% on average, while guaranteeing lower processing latency.},
  doi               = {https://doi.org/10.1109/CloudNet.2018.8549342},
  startyear         = {2018},
  startmonth        = {Oct},
  startday          = {22},
  finishyear        = {2018},
  finishmonth       = {Oct},
  finishday         = {24},
  isbn              = {9781538668313},
  conference        = {2018 IEEE 7th International Conference on Cloud Networking},
  day               = {27},
  publicationstatus = {published},
}

To guarantee an efficient and high-performance environment for mobile devices to perform offloading with low end-to-end delay, it is important to ensure no network policies are violated. In this paper, we explore the simultaneous, dynamic virtual machine (VM) and policy consolidation, and formulate the Policy-VM Latency-aware Consolidation problem for Mobile Edge Computing, which is shown to be NP-Hard. We propose the PL-Edge, an efficient scheme to jointly consolidate network policies and virtual machines for mobile edge computing to reduce communication end-to-end delays among devices and virtual machines. Our simulation results demonstrate that the proposed PL-Edge can significantly reduces policy-flows end-to-end delay by nearly 45% while adhering strictly to the requirements of network policies.

@inproceedings{genez2018latencyawarecomputing,
  author            = {Genez, TAL and Tso, FP and Cui, L},
  booktitle         = {CCNC 2018 - 2018 15th IEEE Annual Consumer Communications and Networking Conference},
  month             = {Mar},
  organization      = {Las Vegas, USA},
  pages             = {1--6},
  title             = {Latency-aware joint virtual machine and policy consolidation for mobile edge computing},
  volume            = {2018-January},
  year              = {2018},
  abstract          = {© 2018 IEEE. To guarantee an efficient and high-performance environment for mobile devices to perform offloading with low end-to-end delay, it is important to ensure no network policies are violated. In this paper, we explore the simultaneous, dynamic virtual machine (VM) and policy consolidation, and formulate the Policy-VM Latency-aware Consolidation problem for Mobile Edge Computing, which is shown to be NP-Hard. We propose the PL-Edge, an efficient scheme to jointly consolidate network policies and virtual machines for mobile edge computing to reduce communication end-to-end delays among devices and virtual machines. Our simulation results demonstrate that the proposed PL-Edge can significantly reduces policy-flows end-to-end delay by nearly 45\% while adhering strictly to the requirements of network policies.},
  doi               = {https://doi.org/10.1109/CCNC.2018.8319204},
  startyear         = {2018},
  startmonth        = {Jan},
  startday          = {12},
  finishyear        = {2018},
  finishmonth       = {Jan},
  finishday         = {15},
  isbn              = {9781538647905},
  conference        = {2018 15th IEEE Annual Consumer Communications and Networking Conference},
  day               = {16},
  publicationstatus = {published},
}

Live migration, underpinned by virtualisation technologies, has enabled improved manageability and fault tolerance for servers. However, virtualised server infrastructures suffer from significant processing overheads, system inconsistencies, security issues and unpredictable performance which makes them unsuitable for low-power and resource-constraint computing devices that processing latency-sensitive, 'Big-data'-type data. Consequently, we ask: 'How do we eliminate the overhead of virtualisation whilst still retaining its benefits?' Motivated by this question, we investigate a practical approach for a bare-metal live migration scheme for ARM-based instances low-power servers and edge devices. In this paper, we position ARM-based bare-metal live migration as a technique that will underpin the efficiency on edge-computing and on Micro-datacentres. We also introduce our early work on identifying three key technical challenges and discuss their solutions.

@inproceedings{avramidis2018livemicrodatacentres,
  author            = {Avramidis, I and Mackay, M and Tso, FP and Fukai, T and Shinagawa, T},
  booktitle         = {CCNC 2018 - 2018 15th IEEE Annual Consumer Communications and Networking Conference},
  month             = {Mar},
  organization      = {Las Vegas, USA},
  pages             = {1--6},
  title             = {Live migration on ARM-based micro-datacentres},
  volume            = {2018-January},
  year              = {2018},
  abstract          = {© 2018 IEEE. Live migration, underpinned by virtualisation technologies, has enabled improved manageability and fault tolerance for servers. However, virtualised server infrastructures suffer from significant processing overheads, system inconsistencies, security issues and unpredictable performance which makes them unsuitable for low-power and resource-constraint computing devices that processing latency-sensitive, 'Big-data'-type data. Consequently, we ask: 'How do we eliminate the overhead of virtualisation whilst still retaining its benefits?' Motivated by this question, we investigate a practical approach for a bare-metal live migration scheme for ARM-based instances low-power servers and edge devices. In this paper, we position ARM-based bare-metal live migration as a technique that will underpin the efficiency on edge-computing and on Micro-datacentres. We also introduce our early work on identifying three key technical challenges and discuss their solutions.},
  doi               = {https://doi.org/10.1109/CCNC.2018.8319241},
  startyear         = {2018},
  startmonth        = {Jan},
  startday          = {12},
  finishyear        = {2018},
  finishmonth       = {Jan},
  finishday         = {15},
  isbn              = {9781538647905},
  conference        = {2018 15th IEEE Annual Consumer Communications \& Networking Conference (CCNC)},
  day               = {16},
  publicationstatus = {published},
}

As a vital component of TCP, congestion control defines TCP's performance characteristics. Hence, it is important for congestion control to provide high link utilization and low queuing delay. Recent BBR tries to estimate available bottleneck capacity to achieve this goal. However, its aggressiveness characteristics generate a massive amount of packet retransmission which harms loss-based congestion control protocol such as Cubic. In this paper, we first dive into this issue and reveal that the aggressiveness of BBR can degrade the performance of Cubic, as well as the overall Internet transmission. Then we present Modest BBR, a simple yet effective solution based on BBR, by responding to retransmission less aggressively. Through extensive testbed experiments and Mininet simulation, we show Modest BBR can preserve high throughput and short convergence time while improve the overall performance when coexisting with Cubic. For example, Modest BBR gets similar throughput compared to BBR, while it improves 7.1% of the overall throughput and achieves better fairness to loss-based schemes.

@inproceedings{zhang2018modestcontrol,
  author            = {Zhang, Y and Cui, L and Tso, FP},
  booktitle         = {Proceedings - IEEE Symposium on Computers and Communications},
  month             = {Nov},
  organization      = {Natal, Brazil},
  pages             = {646--651},
  title             = {Modest BBR: Enabling Better Fairness for BBR Congestion Control},
  volume            = {2018-June},
  year              = {2018},
  abstract          = {© 2018 IEEE. As a vital component of TCP, congestion control defines TCP's performance characteristics. Hence, it is important for congestion control to provide high link utilization and low queuing delay. Recent BBR tries to estimate available bottleneck capacity to achieve this goal. However, its aggressiveness characteristics generate a massive amount of packet retransmission which harms loss-based congestion control protocol such as Cubic. In this paper, we first dive into this issue and reveal that the aggressiveness of BBR can degrade the performance of Cubic, as well as the overall Internet transmission. Then we present Modest BBR, a simple yet effective solution based on BBR, by responding to retransmission less aggressively. Through extensive testbed experiments and Mininet simulation, we show Modest BBR can preserve high throughput and short convergence time while improve the overall performance when coexisting with Cubic. For example, Modest BBR gets similar throughput compared to BBR, while it improves 7.1\% of the overall throughput and achieves better fairness to loss-based schemes.},
  doi               = {https://doi.org/10.1109/ISCC.2018.8538521},
  startyear         = {2018},
  startmonth        = {Jun},
  startday          = {25},
  finishyear        = {2018},
  finishmonth       = {Jun},
  finishday         = {28},
  isbn              = {9781538669501},
  issn              = {1530-1346},
  conference        = {IEEE Symposium on Computers and Communications},
  day               = {15},
  publicationstatus = {published},
}

Until recently, cluster computing was too expensive and too complex for commodity users. However the phenomenal popularity of single board computers like the Raspberry Pi has caused the emergence of the single board computer cluster. This demonstration will present a cheap, practical and portable Raspberry Pi cluster called Pi Stack. We will show pragmatic custom solutions to hardware issues, such as power distribution, and software issues, such as remote updating. We also sketch potential use cases for Pi Stack and other commodity single board computer cluster architectures.

@inproceedings{singer2018nextclusters,
  author            = {Singer, J and Herry, H and Basford, PJ and Hajji, W and Perkins, CS and Tso, FP and Pezaros, D and Mullins, RD and Yoneki, E and Cox, SJ and Johnston, SJ},
  booktitle         = {IEEE/IFIP Network Operations and Management Symposium: Cognitive Management in a Cyber World, NOMS 2018},
  month             = {Jul},
  organization      = {Taipei Taiwan},
  pages             = {1--3},
  title             = {Next generation single board clusters},
  year              = {2018},
  abstract          = {© 2018 IEEE. Until recently, cluster computing was too expensive and too complex for commodity users. However the phenomenal popularity of single board computers like the Raspberry Pi has caused the emergence of the single board computer cluster. This demonstration will present a cheap, practical and portable Raspberry Pi cluster called Pi Stack. We will show pragmatic custom solutions to hardware issues, such as power distribution, and software issues, such as remote updating. We also sketch potential use cases for Pi Stack and other commodity single board computer cluster architectures.},
  doi               = {https://doi.org/10.1109/NOMS.2018.8406120},
  startyear         = {2018},
  startmonth        = {Apr},
  startday          = {23},
  finishyear        = {2018},
  finishmonth       = {Apr},
  finishday         = {27},
  isbn              = {9781538634165},
  conference        = {2018 IEEE/IFIP Network Operations and Management Symposium},
  day               = {6},
  publicationstatus = {published},
}

@inproceedings{tso2018topicwelcome,
  author            = {Tso, F and Trehan, A and Phillips, I and Cui, L},
  booktitle         = {Proceedings of the Annual ACM Symposium on Principles of Distributed Computing},
  month             = {Jul},
  organization      = {Royal Holloway College, University of London},
  pages             = {iii--iii},
  title             = {TOPIC 2018 Chairs' welcome},
  volume            = {Part F138530},
  year              = {2018},
  startyear         = {2018},
  startmonth        = {Jul},
  startday          = {23},
  finishyear        = {2018},
  finishmonth       = {Jul},
  finishday         = {27},
  isbn              = {9781450357760},
  conference        = {Principles of Distributed Computing},
  day               = {23},
  publicationstatus = {published},
}

Cloud computing is revolutionizing many IT ecosystems through offering scalable computing resources that are easy to configure, use and inter-connect. However, this model has always been viewed with some suspicion as it raises a wide range of security and privacy issues that need to be negotiated. This research focuses on the construction of a trust layer in cloud computing to build a trust relationship between cloud service providers and cloud users. In particular, we address the rise of container-based virtualisation has a weak isolation compared to traditional VMs because of the shared use of the OS kernel and system components. Therefore, we will build a trust layer to solve the issues of weaker isolation whilst maintaining the performance and scalability of the approach. This paper has two objectives. Firstly, we propose a security system to protect containers from other guests through the addition of a Role-based Access Control (RBAC) model and the provision of strict data protection and security. Secondly, we provide a stress test using isolation benchmarking tools to evaluate the isolation in containers in term of performance.

@inproceedings{alobaidan2017buildvirtualisation,
  author            = {Alobaidan, I and Mackay, M and Tso, P},
  booktitle         = {Proceedings - 2016 9th International Conference on Developments in eSystems Engineering, DeSE 2016},
  month             = {May},
  pages             = {143--148},
  title             = {Build Trust in the Cloud Computing - Isolation in Container Based Virtualisation},
  year              = {2017},
  abstract          = {Cloud computing is revolutionizing many IT ecosystems through offering scalable computing resources that are easy to configure, use and inter-connect. However, this model has always been viewed with some suspicion as it raises a wide range of security and privacy issues that need to be negotiated. This research focuses on the construction of a trust layer in cloud computing to build a trust relationship between cloud service providers and cloud users. In particular, we address the rise of container-based virtualisation has a weak isolation compared to traditional VMs because of the shared use of the OS kernel and system components. Therefore, we will build a trust layer to solve the issues of weaker isolation whilst maintaining the performance and scalability of the approach. This paper has two objectives. Firstly, we propose a security system to protect containers from other guests through the addition of a Role-based Access Control (RBAC) model and the provision of strict data protection and security. Secondly, we provide a stress test using isolation benchmarking tools to evaluate the isolation in containers in term of performance.},
  doi               = {https://doi.org/10.1109/DeSE.2016.24},
  isbn              = {9781509054879},
  day               = {17},
  publicationstatus = {published},
}

Middleboxes (MBs) are ubiquitous in modern data centre (DC) due to their crucial role in implementing network security, management and optimisation. In order to meet network policy's requirement on correct traversal of an ordered sequence of MBs, network administrators rely on static policy based routing or VLAN stitching to steer traffic flows. However, dynamic virtual server migration in virtual environment has greatly challenged such static traffic steering. In this paper, we design and implement Sync, an efficient and synergistic scheme to jointly consolidate network policies and virtual machines (VMs), in a readily deployable Mininet environment. We present the architecture of Sync framework and open source its code. We also extensively evaluate Sync over diverse workload and policies. Our results show that in an emulated DC of 686 servers, 10k VMs, 8k policies, and 100k flows, Sync processes a group of 900 VMs and 10 VMs in 634 seconds and 4 seconds respectively.

@inproceedings{hajji2017experimentalmachines,
  author            = {Hajji, W and Tso, FP and Cui, L and Pezaros, DP},
  booktitle         = {Proceedings - IEEE Symposium on Computers and Communications},
  month             = {Sep},
  organization      = {Heraklion, Crete, Greece},
  pages             = {1338--1343},
  title             = {Experimental evaluation of SDN-controlled, joint consolidation of policies and virtual machines},
  year              = {2017},
  abstract          = {© 2017 IEEE. Middleboxes (MBs) are ubiquitous in modern data centre (DC) due to their crucial role in implementing network security, management and optimisation. In order to meet network policy's requirement on correct traversal of an ordered sequence of MBs, network administrators rely on static policy based routing or VLAN stitching to steer traffic flows. However, dynamic virtual server migration in virtual environment has greatly challenged such static traffic steering. In this paper, we design and implement Sync, an efficient and synergistic scheme to jointly consolidate network policies and virtual machines (VMs), in a readily deployable Mininet environment. We present the architecture of Sync framework and open source its code. We also extensively evaluate Sync over diverse workload and policies. Our results show that in an emulated DC of 686 servers, 10k VMs, 8k policies, and 100k flows, Sync processes a group of 900 VMs and 10 VMs in 634 seconds and 4 seconds respectively.},
  doi               = {https://doi.org/10.1109/ISCC.2017.8024710},
  startyear         = {2018},
  startmonth        = {Jul},
  startday          = {3},
  finishyear        = {2018},
  finishmonth       = {Jul},
  finishday         = {6},
  isbn              = {9781538616291},
  issn              = {1530-1346},
  conference        = {IEEE Symposium on Computers and Communications},
  day               = {1},
  publicationstatus = {published},
}

With the emergence of network function virtualization, data center start to deploy a variety of network function boxes (NFBs) in both physical and virtual form factors in order to combines inherent efficiency offered by physical NFBs with the agility and flexibility of virtual ones. However, existing schemes are limited to exclusively consider physical or virtual NFBs, which may reduce the performance efficiency of services running atop. In this paper, we propose a Heterogeneous NetwOrk Policy Enforcement scheme (HOPE) to overcome these challenges. An efficient algorithm that can closely approximate optimal latencywise NF service chaining is proposed. The experimental results have also shown that HOPE can outperform greedy algorithm by 25% in terms of network latency and is 56x more efficient than naive depth-first search algorithm.

@inproceedings{cui2017heterogeneouscenters,
  author       = {Cui, L and Tso, FP and Jia, W},
  booktitle    = {},
  month        = {Jan},
  note         = {© 2017 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.},
  organization = {Lisbon, Portugal},
  publisher    = {© IEEE},
  title        = {Heterogeneous network policy enforcement in data centers},
  url          = {https://doi.org/10.23919/INM.2017.7987327},
  year         = {2017},
  abstract     = {With the emergence of network function virtualization, data center start to deploy a variety of network function boxes (NFBs) in both physical and virtual form factors in order to combines inherent efficiency offered by physical NFBs with the agility and flexibility of virtual ones. However, existing schemes are limited to exclusively consider physical or virtual NFBs, which may reduce the performance efficiency of services running
                  atop. In this paper, we propose a Heterogeneous NetwOrk Policy Enforcement scheme (HOPE) to overcome these challenges. An efficient algorithm that can closely approximate optimal latencywise NF service chaining is proposed. The experimental results have also shown that HOPE can outperform greedy algorithm by 25\% in terms of network latency and is 56x more efficient than naive depth-first search algorithm.},
  doi          = {10.23919/INM.2017.7987327},
  isbn         = {9783901882890},
  keyword      = {469999 Other information and computing sciences not elsewhere classified},
  keyword      = {089999 Information and Computing Sciences not elsewhere classified},
  language     = {English},
  conference   = {IFIP/IEEE International Symposium on Integrated Network Management},
  day          = {1},
  version      = {Figshare item version: 1},
}

This paper discusses the use of machine learning techniques for the classification of medical data, specifically for guiding disease modifying therapies for Sickle Cell. Extensive research has indicated that machine learning approaches generate significant improvements when used for the pre-processing of medical time-series data signals and have assisted in obtaining high accuracy in the classification of medical data. The aim of this paper is to present findings for several classes of learning algorithm for medically related problems. The initial case study addressed in this paper involves classifying the dosage of medication required for the treatment of patients with Sickle Cell Disease. We use different machine learning architectures in order to investigate the accuracy and performance within the case study. The main purpose of applying classification approach is to enable healthcare organisations to provide accurate amount of medication. The results obtained from a range of models during our experiments have shown that of the proposed models, recurrent networks produced inferior results in comparison to conventional feedforward neural networks and the Random Forest model. For our dataset, it was found that the Random Forest Classifier produced the highest levels of performance overall.

@article{khalaf2017machinemodels,
  author            = {Khalaf, M and Hussain, AJ and Keight, R and Al-Jumeily, D and Fergus, P and Keenan, R and Tso, P},
  journal           = {Neurocomputing},
  month             = {Mar},
  pages             = {154--164},
  title             = {Machine learning approaches to the application of disease modifying therapy for sickle cell using classification models},
  volume            = {228},
  year              = {2017},
  abstract          = {This paper discusses the use of machine learning techniques for the classification of medical data, specifically for guiding disease modifying therapies for Sickle Cell. Extensive research has indicated that machine learning approaches generate significant improvements when used for the pre-processing of medical time-series data signals and have assisted in obtaining high accuracy in the classification of medical data. The aim of this paper is to present findings for several classes of learning algorithm for medically related problems. The initial case study addressed in this paper involves classifying the dosage of medication required for the treatment of patients with Sickle Cell Disease. We use different machine learning architectures in order to investigate the accuracy and performance within the case study. The main purpose of applying classification approach is to enable healthcare organisations to provide accurate amount of medication. The results obtained from a range of models during our experiments have shown that of the proposed models, recurrent networks produced inferior results in comparison to conventional feedforward neural networks and the Random Forest model. For our dataset, it was found that the Random Forest Classifier produced the highest levels of performance overall.},
  doi               = {https://doi.org/10.1016/j.neucom.2016.10.043},
  issn              = {0925-2312},
  eissn             = {1872-8286},
  day               = {8},
  publicationstatus = {published},
}

In order to minimise their energy use, data centre operators are constantly exploring new ways to construct computing infrastructures. As low power CPUs, exemplified by ARM-based devices, are becoming increasingly popular, there is a growing trend for the large scale deployment of low power servers in data centres. For example, recent research has shown promising results on constructing small scale data centres using Raspberry Pi (RPi) single-board computers as their building blocks. To enable larger scale experimentation and feasibility studies, cloud simulators could be utilised. Unfortunately, stateof-the-art simulators often need significant modification to include such low power devices as core data centre components. In this paper, we introduce models and extensions to estimate the behaviour of these new components in the DISSECT-CF cloud computing simulator. We show that how a RPi based cloud could be simulated with the use of the new models. We evaluate the precision and behaviour of the implemented models using a Hadoop-based application scenario executed both in real life and simulated clouds.

@inproceedings{kecskemeti2017modellingsimulation,
  author       = {Kecskemeti, G and Hajji, W and Tso, FP},
  booktitle    = {},
  month        = {Jan},
  note         = {© 2017 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.},
  organization = {St. Petersburg, Russia},
  publisher    = {© IEEE},
  title        = {Modelling low power compute clusters for cloud simulation},
  url          = {https://doi.org/10.1109/PDP.2017.33},
  year         = {2017},
  abstract     = {In order to minimise their energy use, data centre operators are constantly exploring new ways to construct computing infrastructures. As low power CPUs, exemplified by ARM-based devices, are becoming increasingly popular, there is a growing trend for the large scale deployment of low power servers in data centres. For example, recent research has shown promising results on constructing small scale data centres using Raspberry Pi (RPi) single-board computers as their building blocks. To enable larger scale experimentation and feasibility studies, cloud simulators could be utilised. Unfortunately, stateof-the-art simulators often need significant modification to include such low power devices as core data centre components. In this paper, we introduce models and extensions to estimate the behaviour of these new components in the DISSECT-CF cloud computing simulator. We show that how a RPi based
                  cloud could be simulated with the use of the new models. We evaluate the precision and behaviour of the implemented models using a Hadoop-based application scenario executed both in real life and simulated clouds.},
  doi          = {https://doi.org/10.1109/PDP.2017.33},
  keyword      = {469999 Other information and computing sciences not elsewhere classified},
  keyword      = {Cloud computing},
  keyword      = {IaaS},
  keyword      = {Low power},
  keyword      = {Simulation},
  keyword      = {Containers},
  keyword      = {Docker},
  keyword      = {089999 Information and Computing Sciences not elsewhere classified},
  language     = {English},
  conference   = {The 25th Euromicro International Conference on Parallel, Distributed, and Network-Based Processing (PDP 2017)},
  day          = {1},
  version      = {Figshare item version: 1},
}

Virtualization is the underpinning technology enabling cloud computing service provisioning, and container-based virtualization provides an efficient sharing of the underlying host kernel libraries amongst multiple guests. While there has been research on protecting the host against compromise by malicious guests, research on protecting the guests against a compromised host is limited. In this paper, we present an access control solution which prevents the host from gaining access into the guest containers and their data. Using system call interception together with the built-in AppArmor mandatory access control (MAC) approach the solution protects guest containers from a malicious host attempting to compromise the integrity of data stored therein. Evaluation of results have shown that it can effectively prevent hostile access from host to guest containers while ensuring minimal performance overhead.

@inproceedings{win2017protectinterception,
  author            = {Win, TY and Tso, FP and Mair, Q and Tianfield, H},
  booktitle         = {Proceedings - 14th International Symposium on Pervasive Systems, Algorithms and Networks, I-SPAN 2017, 11th International Conference on Frontier of Computer Science and Technology, FCST 2017 and 3rd International Symposium of Creative Computing, ISCC 2017},
  month             = {Nov},
  organization      = {Exeter, United Kingdom},
  pages             = {191--196},
  title             = {PROTECT: Container process isolation using system call interception},
  year              = {2017},
  abstract          = {© 2017 IEEE. Virtualization is the underpinning technology enabling cloud computing service provisioning, and container-based virtualization provides an efficient sharing of the underlying host kernel libraries amongst multiple guests. While there has been research on protecting the host against compromise by malicious guests, research on protecting the guests against a compromised host is limited. In this paper, we present an access control solution which prevents the host from gaining access into the guest containers and their data. Using system call interception together with the built-in AppArmor mandatory access control (MAC) approach the solution protects guest containers from a malicious host attempting to compromise the integrity of data stored therein. Evaluation of results have shown that it can effectively prevent hostile access from host to guest containers while ensuring minimal performance overhead.},
  doi               = {https://doi.org/10.1109/ISPAN-FCST-ISCC.2017.24},
  startyear         = {2017},
  startmonth        = {Jun},
  startday          = {21},
  finishyear        = {2017},
  finishmonth       = {Jun},
  finishday         = {23},
  isbn              = {9781538608401},
  conference        = {2017 14th International Symposium on Pervasive Systems, Algorithms and Networks \& 2017 11th International Conference on Frontier of Computer Science and Technology \& 2017 Third International Symposium of Creative Computing (ISPAN-FCST-ISCC)},
  day               = {27},
  publicationstatus = {published},
}

Nowadays, many web services (e.g., cloud storage) are deployed inside datacenters and may trigger transfers to clients through WAN. TCP congestion control is a vital component for improving the performance (e.g., latency) of these services. Considering complex networking environment, the default congestion control algorithms on servers may not always be the most efficient, and new advanced algorithms will be proposed. However, adjusting congestion control algorithm usually requires modification of TCP stacks of servers, which is difficult if not impossible, especially considering different operating systems and configurations on servers. In this paper, we propose TCon, a light-weight, flexible and scalable platform that allows administrators (or operators) to deploy any appropriate congestion control algorithms transparently without making any changes to TCP stacks of servers. We have implemented TCon in Open vSwitch (OVS) and conducted extensive test-bed experiments by transparently deploying BBR congestion control algorithm over TCon. Test-bed results show that the BBR over TCon works effectively and the performance stays close to its native implementation on servers, reducing latency by 12.76% on average.

@inproceedings{zhang2017tcontransfers,
  author       = {Zhang, Y and Cui, L and Tso, FP and Guan, Q and Jia, W},
  booktitle    = {IFIP International Conference on Network and Parallel Computing (NPC 2017)},
  month        = {Jan},
  note         = {This is a pre-copyedited version of a contribution published in Shi X. ... et al (eds). Network and Parallel Computing published by Springer. The definitive authenticated version is available online via http://dx.doi.org/10.1007/978-3-319-68210-5_5. This paper was also presented at the IFIP International Conference on Network and Parallel Computing (NPC 2017), Hefei, China, 20th-21st October 2017.},
  organization = {Hefei, China},
  pages        = {49--61},
  publisher    = {Springer © IFIP International Federation for Information Processing},
  series       = {Lecture Notes in Computer Science;10578},
  title        = {TCon: A transparent congestion control deployment platform for optimizing WAN transfers},
  url          = {https://doi.org/10.1007/978-3-319-68210-5_5},
  volume       = {10578 LNCS},
  year         = {2017},
  abstract     = {Nowadays, many web services (e.g., cloud storage) are deployed inside datacenters and may trigger transfers to clients through WAN. TCP congestion control is a vital component for improving the performance (e.g., latency) of these services. Considering complex networking environment, the default congestion control algorithms on servers may not always be the most efficient, and new advanced algorithms will be proposed. However, adjusting congestion control algorithm usually requires modification of TCP stacks of servers, which is difficult if not impossible, especially considering different operating systems and configurations on servers. In this paper, we propose TCon, a light-weight, flexible and scalable platform that allows administrators (or operators) to deploy any appropriate congestion control algorithms transparently without making any changes to TCP stacks of servers. We have implemented TCon in Open vSwitch (OVS) and conducted extensive test-bed experiments by transparently deploying BBR congestion control algorithm over TCon. Test-bed results show that the BBR over TCon works effectively and the performance stays close to its native implementation on servers, reducing latency by 12.76\% on average.},
  doi          = {https://doi.org/10.1007/978-3-319-68210-5_5},
  isbn         = {9783319682099},
  issn         = {0302-9743},
  keyword      = {469999 Other information and computing sciences not elsewhere classified},
  keyword      = {Congestion control},
  keyword      = {BBR},
  keyword      = {Transparent},
  keyword      = {089999 Information and Computing Sciences not elsewhere classified},
  language     = {English},
  day          = {1},
  version      = {Figshare item version: 1},
}

The centralization of control plane in Software defined networking (SDN) creates a paramount challenge on troubleshooting the network as packets are ultimately forwarded by distributed data planes. Existing path tracing tools largely utilize packet tags to probe network paths among SDN-enabled switches. However, network functions (NFs) or middleboxes, whose presence is ubiquitous in today's networks, can drop packets or alter their tags - an action that can collapse the probing mechanism. In addition, sending probing packets through network functions could corrupt their internal states, risking of the correctness of servicing logic (e.g., incorrect load balancing decisions). In this paper, we present a novel troubleshooting tool, Track, for SDN-enabled network with arbitrary NFs. Track can discover the forwarding path including NFs taken by any packets, without changing the forwarding rules in switches and internal states of NFs. We have implemented Track on RYU controller. Our extensive experiment results show that Track can achieve 95.08% and 100% accuracy for discovering forwarding paths with and without NFs respectively, and can efficiently generate traces within 3 milliseconds per hop.

@inproceedings{zhang2017trackfunctions,
  author            = {Zhang, Y and Cui, L and Tso, FP and Zhang, Y},
  booktitle         = {Proceedings of the 2017 IEEE 6th International Conference on Cloud Networking, CloudNet 2017},
  month             = {Oct},
  organization      = {Prague, Czech Republic},
  title             = {Track: Tracerouting in SDN networks with arbitrary network functions},
  year              = {2017},
  abstract          = {© 2017 IEEE. The centralization of control plane in Software defined networking (SDN) creates a paramount challenge on troubleshooting the network as packets are ultimately forwarded by distributed data planes. Existing path tracing tools largely utilize packet tags to probe network paths among SDN-enabled switches. However, network functions (NFs) or middleboxes, whose presence is ubiquitous in today's networks, can drop packets or alter their tags - an action that can collapse the probing mechanism. In addition, sending probing packets through network functions could corrupt their internal states, risking of the correctness of servicing logic (e.g., incorrect load balancing decisions). In this paper, we present a novel troubleshooting tool, Track, for SDN-enabled network with arbitrary NFs. Track can discover the forwarding path including NFs taken by any packets, without changing the forwarding rules in switches and internal states of NFs. We have implemented Track on RYU controller. Our extensive experiment results show that Track can achieve 95.08\% and 100\% accuracy for discovering forwarding paths with and without NFs respectively, and can efficiently generate traces within 3 milliseconds per hop.},
  doi               = {https://doi.org/10.1109/CloudNet.2017.8071526},
  startyear         = {2018},
  startmonth        = {Sep},
  startday          = {25},
  finishyear        = {2018},
  finishmonth       = {Sep},
  finishday         = {27},
  isbn              = {9781509040261},
  conference        = {The 6th IEEE International Conference on Cloud Networking (IEEE CloudNet 2017)},
  day               = {18},
  publicationstatus = {published},
}

Proliferation of the number of smart devices and user applications has generated a tremendous volume of data traffic from/to a cellular network. With a traditional cellular network, a user may experience many drawbacks such as low throughput, large latencies and network outages due to overload of data traffic. The software defined network (SDN) and network function virtualization (NFV) rise as a promising solution to overcome such issues of traditional network architecture. In this paper, we introduce a new network architecture for LTE and WiFi slicing networks taking into account the advantage of SDN and NFV concepts. We propose an IPFlow management controller in a slicing network to offload and balance the flow data traffic. By utilizing the P-GW and Wireless Access Gateway, we can handle the IP-Flow between LTE and WiFi networks. The P-GW works as an IP-Flow anchor to maintain the flow seamlessly during the offloading and balancing IP-Flow. Within WiFi networks, we leverage the Light Virtual Access Point (LVAP) approach to abstract the WiFi protocol stack for a programming capability of centralized control of WiFi network through the WiFi controller. By creating a client virtual port and assigning a specific Service Set Identifier (SSID), we give a capability to slice an operator’s network to control over his clients within a WiFi coverage area network.

@inproceedings{alfoudi2017trafficnetworks,
  author       = {Alfoudi, AS and Dighriri, M and Lee, GM and Pereira, R and Tso, FP},
  booktitle    = {},
  month        = {Jan},
  note         = {© 2017 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.},
  organization = {Paris, France},
  publisher    = {© IEEE},
  title        = {Traffic management in LTE-WiFi slicing networks},
  url          = {https://doi.org/10.1109/ICIN.2017.7899423},
  year         = {2017},
  abstract     = {Proliferation of the number of smart devices and user applications has generated a tremendous volume of data traffic
                  from/to a cellular network. With a traditional cellular network, a user may experience many drawbacks such as low throughput, large latencies and network outages due to overload of data traffic. The software defined network (SDN) and network function virtualization (NFV) rise as a promising solution to overcome such issues of traditional network architecture. In this paper, we introduce a new network architecture for LTE and WiFi slicing networks taking into account the advantage of SDN and NFV concepts. We propose an IPFlow management controller in a slicing network to offload and balance the flow data traffic. By utilizing the P-GW and Wireless Access Gateway, we can handle the IP-Flow between LTE and WiFi networks. The P-GW works as an IP-Flow anchor to maintain the flow seamlessly during the offloading and balancing IP-Flow. Within WiFi networks, we leverage the Light Virtual Access Point (LVAP)
                  approach to abstract the WiFi protocol stack for a programming
                  capability of centralized control of WiFi network through the WiFi controller. By creating a client virtual port and assigning a specific Service Set Identifier (SSID), we give a capability to slice an operator’s network to control over his clients within a WiFi coverage area network.},
  doi          = {https://doi.org/10.1109/ICIN.2017.7899423},
  keyword      = {469999 Other information and computing sciences not elsewhere classified},
  keyword      = {Long-Term Evolution (LTE)},
  keyword      = {WiFi},
  keyword      = {Software Defined Network (SDN)},
  keyword      = {Network Function Virtualization (NFV)},
  keyword      = {Light Virtual Access Point (LVAP)},
  keyword      = {IP-Flow},
  keyword      = {Data offload},
  keyword      = {089999 Information and Computing Sciences not elsewhere classified},
  language     = {English},
  conference   = {2017 20th Conference on Innovations in Clouds, Internet and Networks (ICIN)},
  day          = {1},
  version      = {Figshare item version: 1},
}

The advent of virtualisation and the increasing demand for outsourced, elastic compute charged on a pay-as-you-use basis has stimulated the development of large-scale Cloud Data Centres (DCs) housing tens of thousands of computer clusters. Of the significant capital outlay required for building and operating such infrastructures, server and network equipment account for 45 and 15% of the total cost, respectively, making resource utilisation efficiency paramount in order to increase the operators’ Return-on-Investment (RoI). In this paper, we present an extensive survey on the management of server and network resources over virtualised Cloud DC infrastructures, highlighting key concepts and results, and critically discussing their limitations and implications for future research opportunities. We highlight the need for and benefits of adaptive resource provisioning that alleviates reliance on static utilisation prediction models and exploits direct measurement of resource utilisation on servers and network nodes. Coupling such distributed measurement with logically centralised Software Defined Networking (SDN) principles, we subsequently discuss the challenges and opportunities for converged resource management over converged ICT environments, through unifying control loops to globally orchestrate adaptive and load-sensitive resource provisioning.

@article{tso2016networksurvey,
  author            = {Tso, FP and Jouet, S and Pezaros, DP},
  journal           = {Computer Networks},
  month             = {Sep},
  pages             = {209--225},
  title             = {Network and server resource management strategies for data centre infrastructures: A survey},
  volume            = {106},
  year              = {2016},
  abstract          = {The advent of virtualisation and the increasing demand for outsourced, elastic compute charged on a pay-as-you-use basis has stimulated the development of large-scale Cloud Data Centres (DCs) housing tens of thousands of computer clusters. Of the significant capital outlay required for building and operating such infrastructures, server and network equipment account for 45 and 15\% of the total cost, respectively, making resource utilisation efficiency paramount in order to increase the operators’ Return-on-Investment (RoI). In this paper, we present an extensive survey on the management of server and network resources over virtualised Cloud DC infrastructures, highlighting key concepts and results, and critically discussing their limitations and implications for future research opportunities. We highlight the need for and benefits of adaptive resource provisioning that alleviates reliance on static utilisation prediction models and exploits direct measurement of resource utilisation on servers and network nodes. Coupling such distributed measurement with logically centralised Software Defined Networking (SDN) principles, we subsequently discuss the challenges and opportunities for converged resource management over converged ICT environments, through unifying control loops to globally orchestrate adaptive and load-sensitive resource provisioning.},
  doi               = {https://doi.org/10.1016/j.comnet.2016.07.002},
  issn              = {1389-1286},
  day               = {4},
  publicationstatus = {published},
}

Policies play an important role in network configuration and therefore in offering secure and high performance services especially over multi-tenant Cloud Data Center (DC) environments. At the same time, elastic resource provisioning through virtualization often disregards policy requirements, assuming that the policy implementation is handled by the underlying network infrastructure. This can result in policy violations, performance degradation and security vulnerabilities. In this paper, we define PLAN, a PoLicy-Aware and Network-aware VM management scheme to jointly consider DC communication cost reduction through Virtual Machine (VM) migration while meeting network policy requirements. We show that the problem is NP-hard and derive an efficient approximate algorithm to reduce communication cost while adhering to policy constraints. Through extensive evaluation, we show that PLAN can reduce topology-wide communication cost by 38% over diverse aggregate traffic and configuration policies.

@article{cui2016plancenters,
  author            = {Cui, L and Tso, FP and Pezaros, D and Jia, W and Zhao, W},
  journal           = {IEEE Transactions on Parallel and Distributed Systems},
  month             = {Jan},
  number            = {99},
  title             = {PLAN: Joint policy- and network-aware VM management for cloud data centers},
  volume            = {PP},
  year              = {2016},
  abstract          = {© 1990-2012 IEEE. Policies play an important role in network configuration and therefore in offering secure and high performance services especially over multi-tenant Cloud Data Center (DC) environments. At the same time, elastic resource provisioning through virtualization often disregards policy requirements, assuming that the policy implementation is handled by the underlying network infrastructure. This can result in policy violations, performance degradation and security vulnerabilities. In this paper, we define PLAN, a PoLicy-Aware and Network-aware VM management scheme to jointly consider DC communication cost reduction through Virtual Machine (VM) migration while meeting network policy requirements. We show that the problem is NP-hard and derive an efficient approximate algorithm to reduce communication cost while adhering to policy constraints. Through extensive evaluation, we show that PLAN can reduce topology-wide communication cost by 38\% over diverse aggregate traffic and configuration policies.},
  doi               = {https://doi.org/10.1109/TPDS.2016.2604811},
  issn              = {1045-9219},
  day               = {1},
  publicationstatus = {published},
}

@article{cziva2016sdnbasedcenters,
  author            = {Cziva, R and Jouet, S and Stapleton, D and Tso, FP and Pezaros, DP},
  journal           = {IEEE Transactions on Network and Service Management},
  month             = {Jun},
  number            = {2},
  pages             = {212--225},
  publisher         = {Institute of Electrical and Electronics Engineers (IEEE)},
  title             = {SDN-Based Virtual Machine Management for Cloud Data Centers},
  url               = {http://dx.doi.org/10.1109/tnsm.2016.2528220},
  volume            = {13},
  year              = {2016},
  doi               = {https://doi.org/10.1109/tnsm.2016.2528220},
  eissn             = {1932-4537},
  publicationstatus = {published},
}

In modern Cloud Data Centers (DC)s, correct implementation of network policies is crucial to provide secure, efficient and high performance services for tenants. It is reported that the inefficient management of network policies accounts for 78% of DC downtime, challenged by the dynamically changing network characteristics and by the effects of dynamic Virtual Machine (VM) consolidation. While there has been significant research in policy and VM management, they have so far been treated as disjoint research problems. In this paper, we explore the simultaneous, dynamic VM and policy consolidation, and formulate the Policy-VM Consolidation (PVC) problem, which is shown to be NP-Hard. We then propose Sync, an efficient and synergistic scheme to jointly consolidate network policies and virtual machines. Extensive evaluation results and a testbed implementation of our controller show that policy and VM migration under Sync significantly reduces flow end-to-end delay by nearly 40%, and network-wide communication cost by 50% within few seconds, while adhering strictly to the requirements of network policies.

@inproceedings{cui2016synergisticcenters,
  author     = {Cui, L and Cziva, R and Tso, FP and Pezaros, DP},
  booktitle  = {},
  month      = {Jul},
  note       = {© 2016 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.},
  pages      = {1--9},
  publisher  = {© IEEE},
  title      = {Synergistic policy and virtual machine consolidation in cloud data centers},
  url        = {https://doi.org/10.1109/INFOCOM.2016.7524354},
  year       = {2016},
  abstract   = {In modern Cloud Data Centers (DC)s, correct implementation of network policies is crucial to provide secure, efficient and high performance services for tenants. It is reported that the inefficient management of network policies accounts for 78\% of DC downtime, challenged by the dynamically changing network characteristics and by the effects of dynamic Virtual Machine (VM) consolidation. While there has been significant research in policy and VM management, they have so far been treated as disjoint research problems. In this paper, we explore the simultaneous, dynamic VM and policy consolidation, and formulate the Policy-VM Consolidation (PVC) problem, which is shown to be NP-Hard. We then propose Sync, an efficient and synergistic scheme to jointly consolidate network policies and virtual machines. Extensive evaluation results and a testbed implementation of our controller show that policy and VM migration under Sync significantly reduces flow end-to-end delay by nearly 40\%, and network-wide communication cost by 50\% within few seconds, while adhering strictly to the requirements of network policies.},
  doi        = {https://doi.org/10.1109/INFOCOM.2016.7524354},
  isbn       = {9781467399531},
  keyword    = {469999 Other information and computing sciences not elsewhere classified},
  keyword    = {089999 Information and Computing Sciences not elsewhere classified},
  language   = {English},
  conference = {The 35th Annual IEEE International Conference on Computer Communications (INFOCOM)},
  day        = {28},
  version    = {Figshare item version: 1},
}

@techreport{tso2016understandingcentres,
  author      = {Tso, FP},
  institution = {Department of Computer Science, Liverpool John Moores University},
  note        = {date: 2016-03-08
                 keywords: publications
                 pubstate: published
                 tppubtype: techreport
                 },
  title       = {Understanding The Network I/O Performance of Heterogenous Virtualisation in Cloud Data Centres},
  url         = {http://www.poscotso.com/wp-content/uploads/2016/03/tso2016vmperf.pdf},
  year        = {2016},
}

Nowadays, Internet-of-Things (IoT) devices generate data at high speed and large volume. Often the data require real-time processing to support high system responsiveness which can be supported by localised Cloud and/or Fog computing paradigms. However, there are considerably large deployments of IoT such as sensor networks in remote areas where Internet connectivity is sparse, challenging the localised Cloud and/or Fog computing paradigms. With the advent of the Raspberry Pi, a credit card-sized single board computer, there is a great opportunity to construct low-cost, low-power portable cloud to support real-time data processing next to IoT deployments. In this paper, we extend our previous work on constructing Raspberry Pi Cloud to study its feasibility for real-time big data analytics under realistic application-level workload in both native and virtualised environments. We have extensively tested the performance of a single node Raspberry Pi 2 Model B with httperf and a cluster of 12 nodes with Apache Spark and HDFS (Hadoop Distributed File System). Our results have demonstrated that our portable cloud is useful for supporting real-time big data analytics. On the other hand, our results have also unveiled that overhead for CPU-bound workload in virtualised environment is surprisingly high, at 67.2%. We have found that, for big data applications, the virtualisation overhead is fractional for small jobs but becomes more significant for large jobs, up to 28.6%.

@article{electronics5020029,
  author         = {Hajji, Wajdi and Tso, Fung Po},
  title          = {Understanding the Performance of Low Power Raspberry Pi Cloud for Big Data},
  journal        = {Electronics},
  volume         = {5},
  year           = {2016},
  number         = {2},
  article-number = {29},
  url            = {https://www.mdpi.com/2079-9292/5/2/29},
  issn           = {2079-9292},
  abstract       = {Nowadays, Internet-of-Things (IoT) devices generate data at high speed and large volume. Often the data require real-time processing to support high system responsiveness which can be supported by localised Cloud and/or Fog computing paradigms. However, there are considerably large deployments of IoT such as sensor networks in remote areas where Internet connectivity is sparse, challenging the localised Cloud and/or Fog computing paradigms. With the advent of the Raspberry Pi, a credit card-sized single board computer, there is a great opportunity to construct low-cost, low-power portable cloud to support real-time data processing next to IoT deployments. In this paper, we extend our previous work on constructing Raspberry Pi Cloud to study its feasibility for real-time big data analytics under realistic application-level workload in both native and virtualised environments. We have extensively tested the performance of a single node Raspberry Pi 2 Model B with httperf and a cluster of 12 nodes with Apache Spark and HDFS (Hadoop Distributed File System). Our results have demonstrated that our portable cloud is useful for supporting real-time big data analytics. On the other hand, our results have also unveiled that overhead for CPU-bound workload in virtualised environment is surprisingly high, at 67.2%. We have found that, for big data applications, the virtualisation overhead is fractional for small jobs but becomes more significant for large jobs, up to 28.6%.},
  doi            = {10.3390/electronics5020029},
}

@inproceedings{cui2015jointcenters,
  author    = {Cui, L and Tso, FP},
  booktitle = {IEEE International Conference on Cloud Networking (CLOUDNET)},
  note      = {date: 2015-10-05
               keywords: publications
               pubstate: published
               tppubtype: conference
               },
  title     = {Joint Virtual Machine and Network Policy Consolidation in Cloud Data Centers},
  url       = {http://www.poscotso.com/wp-content/uploads/2016/03/joint_cui.pdf},
  year      = {2015},
}

@inproceedings{cui2015plancentres,
  author    = {Cui, L and Tso, FP and Pezaros, DP and Jia, W},
  booktitle = {IEEE/ACM Utility \& Cloud Computing (UCC)},
  note      = {Acceptance Ratio: 27.5\%
               
               date: 2015-12-09
               keywords: policy
               pubstate: published
               tppubtype: conference
               },
  title     = {PLAN: A Policy-Aware VM Management Scheme for Cloud Data Centres},
  url       = {http://www.poscotso.com/wp-content/uploads/2016/03/plan_cui.pdf},
  year      = {2015},
}

@inproceedings{cui2015policyawarenetworks,
  author    = {Cui, L and Tso, FP and Pezaros, DP and Jia, W and Zho, W},
  booktitle = {IEEE International Conference on Distributed Computing Systems (ICDCS)},
  note      = {date: 2015-06-08
               keywords: publications
               pubstate: published
               tppubtype: conference
               },
  title     = {Policy-Aware Virtual Machine Management in Data Center Networks},
  url       = {http://www.poscotso.com/wp-content/uploads/2016/03/datacenter_policy.pdf},
  year      = {2015},
}

@patent{jia2014patent,
  author  = {Jia, Weijia and Tso, Fung Po},
  title   = {Mobile internet service system for long distance trains},
  number  = {US 8,737,225 B2},
  year    = {2014},
  url     = {https://patents.google.com/patent/US8737225B2},
}

Virtual Machine (VM) management is a powerful mechanism for providing elastic services over Cloud Data Centers (DC)s. At the same time, the resulting network congestion has been repeatedly reported as the main bottleneck in DCs, even when the overall resource utilization of the infrastructure remains low. However, most current VM management strategies are traffic-agnostic, while the few that are traffic-aware only concern a static initial allocation, ignore bandwidth oversubscription, or do not scale. In this paper we present S-CORE, a scalable VM migration algorithm to dynamically reallocate VMs to servers while minimizing the overall communication footprint of active traffic flows. We formulate the aggregate VM communication as an optimization problem and we then define a novel distributed migration scheme that iteratively adapts to dynamic traffic changes. Through extensive simulation and implementation results, we show that S-CORE achieves significant (up to 87%) communication cost reduction while incurring minimal overhead and downtime.

@inproceedings{tso2014scalablecenters,
  author     = {Tso, FP and Oikonomou, K and Kavvadia, E and Pezaros, DP},
  booktitle  = {},
  month      = {Sep},
  note       = {© 2016 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.},
  publisher  = {© IEEE},
  title      = {Scalable traffic-aware virtual machine management for cloud data centers},
  url        = {https://doi.org/10.1109/ICDCS.2014.32},
  year       = {2014},
  abstract   = {Virtual Machine (VM) management is a powerful mechanism for providing elastic services over Cloud Data Centers (DC)s. At the same time, the resulting network congestion has been repeatedly reported as the main bottleneck in DCs, even when the overall resource utilization of the infrastructure remains low. However, most current VM management strategies are traffic-agnostic, while the few that are traffic-aware only concern a static initial allocation, ignore bandwidth oversubscription, or do not scale. In this paper we present S-CORE, a scalable VM migration algorithm to dynamically reallocate VMs to servers while minimizing the overall communication footprint of active traffic flows. We formulate the aggregate VM communication as an optimization problem and we then define a novel distributed migration scheme that iteratively adapts to dynamic traffic changes. Through extensive simulation and implementation results, we show that S-CORE achieves significant (up to 87\%) communication cost reduction while incurring minimal overhead and downtime.},
  doi        = {https://doi.org/10.1109/ICDCS.2014.32},
  issn       = {1063-6927},
  keyword    = {469999 Other information and computing sciences not elsewhere classified},
  keyword    = {Virtual machine},
  keyword    = {160303 Migration},
  keyword    = {Consolidation},
  keyword    = {Communication cost},
  keyword    = {Scalable},
  keyword    = {Traffic-aware},
  keyword    = {Data center network},
  keyword    = {089999 Information and Computing Sciences not elsewhere classified},
  language   = {English},
  conference = {IEEE 34th International Conference on Distributed Computing Systems (ICDCS)},
  day        = {1},
  version    = {Figshare item version: 1},
}

@inproceedings{cziva2014sdnbasedcenters,
  award     = {Best Paper Award},
  author    = {Cziva, R and Stapleton, D and Tso, FP and Pezaros, DP},
  booktitle = {IEEE International Conference on Cloud Networking (CLOUDNET)},
  note      = {Best Paper Award
               
               date: 2014-10-10
               keywords: publications
               pubstate: published
               tppubtype: conference
               },
  title     = {SDN-based Virtual Machine Management for Cloud Data Centers},
  url       = {http://poscotso.info/wp-content/uploads/2016/03/cloudnet14_cziva.pdf},
  year      = {2014},
}

@inproceedings{tso2013baatdaatcenters,
  author    = {Tso, FP and Pezaros, D},
  booktitle = {IEEE symposium on Computers and Communications (ISCC’13)},
  note      = {date: 2013-04-08
               keywords: publications
               pubstate: published
               tppubtype: conference
               },
  title     = {Baatdaat: Measurement-Based Flow Scheduling for Cloud Data Centers},
  url       = {http://www.poscotso.com/wp-content/uploads/2016/03/tso2013baatdaat.pdf},
  year      = {2013},
}

ABSTRACTRecently, the direct sequence spread spectrum (DSSS)‐based technique has been proposed to trace anonymous network flows. In this technique, homogeneous pseudo‐noise (PN) codes are used to modulate multiple bit signals that are embedded into the target flow as watermarks. This technique could be maliciously used to degrade an anonymous communication network. In this paper, we propose an effective single flow‐based scheme to detect the existence of these watermarks. Our investigation shows that, even if we have no knowledge of the applied PN code, we are still able to detect malicious DSSS watermarks via mean‐square autocorrelation (MSAC) of a single modulated flow's traffic rate time series. MSAC shows periodic peaks because of self‐similarity in the modulated traffic caused by homogeneous PN codes that are used in modulating multiple bit signals. Our scheme has low complexity and does not require any PN code synchronization. We evaluate this detection scheme's effectiveness via simulations. Our results demonstrate a high detection rate with a low false positive rate. Real‐world experiments on Tor also validate the feasibility of the detection scheme. Our scheme is more flexible and accurate than the existing multiflow‐based approach in DSSS watermark detection. We also present a theory for reconstructing the DSSS code once the DSSS code length is known and simulations validate the feasibility. Copyright © 2012 John Wiley & Sons, Ltd.

@article{jia2013blindwatermarks,
  author            = {Jia, W and Tso, FP and Ling, Z and Fu, X and Xuan, D and Yu, W},
  journal           = {Security and Communication Networks},
  month             = {Mar},
  number            = {3},
  pages             = {257--274},
  publisher         = {Wiley},
  title             = {Blind detection of spread spectrum flow watermarks},
  url               = {http://dx.doi.org/10.1002/sec.540},
  volume            = {6},
  year              = {2013},
  abstract          = {<jats:title>ABSTRACT</jats:title><jats:p>Recently, the direct sequence spread spectrum (DSSS)‐based technique has been proposed to trace anonymous network flows. In this technique, homogeneous pseudo‐noise (PN) codes are used to modulate multiple bit signals that are embedded into the target flow as watermarks. This technique could be maliciously used to degrade an anonymous communication network. In this paper, we propose an effective single flow‐based scheme to detect the existence of these watermarks. Our investigation shows that, even if we have no knowledge of the applied PN code, we are still able to detect malicious DSSS watermarks via mean‐square autocorrelation (MSAC) of a single modulated flow's traffic rate time series. MSAC shows periodic peaks because of self‐similarity in the modulated traffic caused by homogeneous PN codes that are used in modulating multiple bit signals. Our scheme has low complexity and does not require any PN code synchronization. We evaluate this detection scheme's effectiveness via simulations. Our results demonstrate a high detection rate with a low false positive rate. Real‐world experiments on Tor also validate the feasibility of the detection scheme. Our scheme is more flexible and accurate than the existing multiflow‐based approach in DSSS watermark detection. We also present a theory for reconstructing the DSSS code once the DSSS code length is known and simulations validate the feasibility. Copyright © 2012 John Wiley \&amp; Sons, Ltd.</jats:p>},
  doi               = {https://doi.org/10.1002/sec.540},
  issn              = {1939-0114},
  eissn             = {1939-0122},
  language          = {en},
  publicationstatus = {published},
}

@article{tso2013dragonnettrains,
  author  = {Tso, FP and Cui, L and Zhang, L and Jia, W and Yao, D and Teng, J and Xuan, D},
  journal = {IEEE Transactions on Mobile Computing},
  note    = {date: 2013-09-01
             keywords: publications
             pubstate: published
             tppubtype: article
             },
  title   = {DragonNet: A Robust Mobile Internet Service System for Long-Distance Trains},
  url     = {http://www.poscotso.com/wp-content/uploads/2016/03/dragonnet-tmc-tso.pdf},
  year    = {2013},
  doi     = {https://doi.org/10.1109/TMC.2012.191},
  issn    = {1536-1233},
}

@inproceedings{tso2013implementingcenters,
  author    = {Tso, FP and Hamilton, G and Oikonomou, K and Pezaros, DP},
  booktitle = {IEEE International Conference on Cloud Computing (CLOUD)},
  note      = {date: 2013-06-10
               keywords: publications
               pubstate: published
               tppubtype: conference
               },
  title     = {Implementing Scalable, Network-Aware Virtual Machine Migration for Cloud Data Centers},
  url       = {http://www.poscotso.com/wp-content/uploads/2016/03/cloud2013-tso.pdf},
  year      = {2013},
}

@article{tso2013improvingengineering,
  author            = {Tso, FP and Pezaros, DP},
  journal           = {IEEE Transactions on Parallel and Distributed Systems},
  month             = {Jun},
  number            = {6},
  pages             = {1139--1148},
  publisher         = {Institute of Electrical and Electronics Engineers (IEEE)},
  title             = {Improving Data Center Network Utilization Using Near-Optimal Traffic Engineering},
  url               = {http://dx.doi.org/10.1109/tpds.2012.343},
  volume            = {24},
  year              = {2013},
  doi               = {https://doi.org/10.1109/tpds.2012.343},
  issn              = {1045-9219},
  publicationstatus = {published},
}

Data Center (DC) networks exhibit much more centralized characteristics than the legacy Internet, yet they are operated by similar distributed routing and control algorithms that fail to exploit topological redundancy to deliver better and more sustainable performance. Multipath protocols, for example, use node-local and heuristic information to only exploit path diversity between shortest paths. In this paper, we use a measurement-based approach to schedule flows over both shortest and non-shortest paths based on temporal network-wide utilization. We present the Baatdaat flow scheduling algorithm which uses spare DC network capacity to mitigate the performance degradation of heavily utilized links. Results show that Baatdaat achieves close to optimal Traffic Engineering by reducing network-wide maximum link utilization by up to 18% over Equal-Cost Multi-Path (ECMP) routing, while at the same time improving flow completion time by 41% - 95%.

@inproceedings{tso2013longernetworks,
  author     = {Tso, FP and Hamilton, G and Weber, R and Perkins, CS and Pezaros, DP},
  booktitle  = {},
  month      = {Jan},
  note       = {This paper is closed access.},
  publisher  = {© IEEE},
  title      = {Longer is better: Exploiting path diversity in data center networks},
  url        = {https://doi.org/10.1109/ICDCS.2013.36},
  year       = {2013},
  abstract   = {Data Center (DC) networks exhibit much more centralized characteristics than the legacy Internet, yet they are operated by similar distributed routing and control algorithms that fail to exploit topological redundancy to deliver better and more sustainable performance. Multipath protocols, for example, use node-local and heuristic information to only exploit path diversity between shortest paths. In this paper, we use a measurement-based approach to schedule flows over both shortest and non-shortest paths based on temporal network-wide utilization. We present the Baatdaat flow scheduling algorithm which uses spare DC network capacity to mitigate the performance degradation of heavily utilized links. Results show that Baatdaat achieves close to optimal Traffic Engineering by reducing network-wide maximum link utilization by up to 18\% over Equal-Cost Multi-Path (ECMP) routing, while at the same time improving flow completion time by 41\% - 95\%.},
  doi        = {https://doi.org/10.1109/ICDCS.2013.36},
  issn       = {1063-6927},
  keyword    = {469999 Other information and computing sciences not elsewhere classified},
  keyword    = {Traffic engineering},
  keyword    = {Multipath scheduling},
  keyword    = {Unequal-cost multipath},
  keyword    = {Data center networks},
  keyword    = {Software defined networking},
  keyword    = {089999 Information and Computing Sciences not elsewhere classified},
  language   = {English},
  conference = {IEEE International Conference on Distributed Computing Systems (ICDCS)},
  day        = {1},
  version    = {Figshare item version: 1},
}

@inproceedings{tso2013theinfrastructures,
  author    = {Tso, FP and White, DR and Jouet, S and Singer, J and Pezaros, D},
  booktitle = {The First International Workshop on Resource Management of Cloud Computing (Co-located with ICDCS 2013)},
  note      = {date: 2013-06-24
               keywords: publications
               pubstate: published
               tppubtype: conference
               },
  title     = {The Glasgow Raspberry Pi Cloud: A Scale Model for Cloud Computing Infrastructures},
  url       = {http://www.poscotso.com/wp-content/uploads/2016/03/the_glasgow_raspberry_pi_cloud_a_scale_model_for_cloud_computing_infrastructures.pdf},
  year      = {2013},
}

@article{tso2012mobilitynetworks,
  author            = {Tso, FP and Teng, J and Jia, W and Xuan, D},
  journal           = {IEEE Transactions on Parallel and Distributed Systems},
  month             = {Oct},
  number            = {10},
  pages             = {1895--1907},
  publisher         = {Institute of Electrical and Electronics Engineers (IEEE)},
  title             = {Mobility: A Double-Edged Sword for HSPA Networks: A Large-Scale Test on Hong Kong Mobile HSPA Networks},
  url               = {http://dx.doi.org/10.1109/tpds.2011.289},
  volume            = {23},
  year              = {2012},
  doi               = {https://doi.org/10.1109/tpds.2011.289},
  issn              = {1045-9219},
  publicationstatus = {published},
}

@techreport{tso2012scoreenvironments,
  author      = {Tso, FP and Oikonomous, K and Kavvdia, E and Hamilton, G and Pezaros, D},
  institution = {School of Computing Science, University of Glasgow},
  note        = {date: 2012-11-01
                 keywords: publications
                 pubstate: published
                 tppubtype: techreport
                 },
  title       = {S-CORE: Scalable Communication Reduction in Data Center Environments},
  url         = {http://www.poscotso.com/wp-content/uploads/2016/03/migration_techreport.pdf},
  year        = {2012},
}

Measurement and inference in data centers present a set of opportunities and challenges distinct from the Internet domain. Existing toolsets may be perturbed or be mislead by issues related to virtualization. Yet, while equally confronted by scale, data centers are relatively homogenous and symmetric. We believe these may be attributes to be exploited. However, data is required to better evaluate our hypotheses. Therefore, we introduce our efforts to gather data using a single framework from which we can launch tests of our choosing. Our observations reinforce recent claims, but indicate changes in the network. They also reveal additional obfuscations stemming from virtualization. © 2012 Authors.

@inproceedings{twigg2012userleveltomography,
  pdf               = {p101.pdf},
  author            = {Twigg, NA and Fayed, M and Perkins, C and Pezaros, D and Tso, P},
  booktitle         = {SIGCOMM'12 - Proceedings of the ACM SIGCOMM 2012 Conference Applications, Technologies, Architectures, and Protocols for Computer Communication},
  month             = {Sep},
  pages             = {101--102},
  title             = {User-level data center tomography},
  year              = {2012},
  abstract          = {Measurement and inference in data centers present a set of opportunities and challenges distinct from the Internet domain. Existing toolsets may be perturbed or be mislead by issues related to virtualization. Yet, while equally confronted by scale, data centers are relatively homogenous and symmetric. We believe these may be attributes to be exploited. However, data is required to better evaluate our hypotheses. Therefore, we introduce our efforts to gather data using a single framework from which we can launch tests of our choosing. Our observations reinforce recent claims, but indicate changes in the network. They also reveal additional obfuscations stemming from virtualization. © 2012 Authors.},
  doi               = {https://doi.org/10.1145/2342356.2342380},
  isbn              = {9781450314190},
  day               = {26},
  publicationstatus = {published},
}

@article{cui2012wefilabeducation,
  author            = {Cui, L and Tso, FP and Yao, D and Jia, W},
  journal           = {IEEE Transactions on Learning Technologies},
  month             = {Oct},
  number            = {4},
  pages             = {291--303},
  publisher         = {Institute of Electrical and Electronics Engineers (IEEE)},
  title             = {WeFiLab: A Web-Based WiFi Laboratory Platform for Wireless Networking Education},
  url               = {http://dx.doi.org/10.1109/tlt.2012.6},
  volume            = {5},
  year              = {2012},
  doi               = {https://doi.org/10.1109/tlt.2012.6},
  issn              = {1939-1382},
  publicationstatus = {published},
}

@article{tso2011buildingservices,
  author  = {Tso, FP and Cui, L and Zhang, L and Jia, W},
  journal = {ZTE Communications},
  note    = {date: 2011-03-01
             keywords: publications
             pubstate: published
             tppubtype: article
             },
  title   = {Building a Platform to Bridge Low End Mobile Phones and Cloud Computing Services},
  url     = {http://www.poscotso.com/wp-content/uploads/2016/03/tso2011building.pdf},
  year    = {2011},
  issn    = {1673-5188},
}

@inproceedings{tso2011dragonnettrains,
  author    = {Tso, FP and Cui, L and Zhang, L and Jia, W and Yao, D and Teng, J and Xuan, D},
  booktitle = {IEEE International Conference on Computer Communications (INFOCOM)},
  note      = {Acceptance Ratio: 15.96\%
               
               date: 2011-04-11
               keywords: publications
               pubstate: published
               tppubtype: conference
               },
  title     = {DragonNet: A Robust Mobile Internet Services System for Long Distance Trains},
  url       = {http://www.poscotso.com/wp-content/uploads/2016/03/tso2011dragonnet.pdf},
  year      = {2011},
  doi       = {https://doi.org/10.1109/INFCOM.2011.5935309},
}

@phdthesis{tso2011thesis,
  author = {Tso, Fung Po},
  title  = {Mobile Internet system implementation and measurement},
  school = {City University of Hong Kong},
  year   = {2011},
}

@inproceedings{tso2010mobilitynetworks,
  doi       = {https://doi.org/10.1145/1860093.1860105},
  author    = {Tso, FP and Teng, J and Jia, W and Xuan, D},
  booktitle = {ACM International Symposium on Mobile Ad Hoc Networking and Computing (MobiHoc)},
  note      = {Acceptance Ratio: 16.9\%
               
               date: 2010-09-22
               keywords: publications
               pubstate: published
               tppubtype: conference
               },
  title     = {Mobility: A Double-Edged Sword for HSPA Networks},
  url       = {http://www.poscotso.com/wp-content/uploads/2016/03/10_mobihoc_ttjx.pdf},
  year      = {2010},
}

The 3G-324M is an umbrella standard of the Third Generation Partnership Project (3GPP) for wireless video communications, which was developed to satisfy the stringent requirements of real-time interactive video and audio services. In practice, 3G-324M has been employed in 3G networks today to enable the multimedia services with messaging and streaming. However, the design of the supporting architecture for the unification of the diverse streams with 3G-324M poses lots of challenges. This chapter introduces a new supporting

@incollection{du2009apredictions,
  author            = {Du, Y and Tso, FP and Jia, W},
  booktitle         = {Ubiquitous Multimedia Computing},
  month             = {Jan},
  pages             = {73--126},
  title             = {A hybrid approach to communicate between wsn and umts with video quality measurements and predictions},
  year              = {2009},
  abstract          = {The 3G-324M is an umbrella standard of the Third Generation Partnership Project (3GPP) for wireless video communications, which was developed to satisfy the stringent requirements of real-time interactive video and audio services. In practice, 3G-324M has been employed in 3G networks today to enable the multimedia services with messaging and streaming. However, the design of the supporting architecture for the unification of the diverse streams with 3G-324M poses lots of challenges. This chapter introduces a new supporting},
  doi               = {https://doi.org/10.1201/9781420090512},
  isbn              = {9781420093384},
  day               = {1},
  publicationstatus = {published},
}

@inproceedings{jia2009blindwatermarks,
  author    = {Jia, W and Tso, FP and Fu, X and Lin, Z and Xuan, D and Yu, W},
  booktitle = {IEEE International Conference on Computer Communications (INFOCOM)},
  note      = {Acceptance Ratio: 19.7\%
               
               date: 2009-04-19
               keywords: publications
               pubstate: published
               tppubtype: conference
               },
  title     = {Blind Detection of Spread Spectrum Flow Watermarks},
  url       = {http://www.poscotso.com/wp-content/uploads/2016/03/jia2009blind.pdf},
  year      = {2009},
  doi       = {https://doi.org/10.1109/INFCOM.2009.5062144},
}

@inproceedings{du2008architecturewsn,
  author    = {Du, Y and Tso, FP and Jia, W},
  booktitle = {IEEE International Conference on Ubi-Media Computing},
  note      = {date: 2008-07-31
               keywords: publications
               pubstate: published
               tppubtype: conference
               },
  title     = {Architecture design of video transmission between UMTS and WSN},
  url       = {http://www.poscotso.com/wp-content/uploads/2016/03/du-2008.pdf},
  year      = {2008},
}

@inproceedings{tso2008designcommunication,
  author    = {Tso, FP and Du, Y and Jia, W},
  booktitle = {IEEE Wireless Telecommunications Symposium},
  note      = {date: 2008-04-24
               keywords: publications
               pubstate: published
               tppubtype: conference
               },
  title     = {Design of an Efficient and Robust Multimedia Gateway for Pervasive Communication},
  url       = {http://poscotso.info/wp-content/uploads/2016/03/tso-2008.pdf},
  year      = {2008},
}

@inproceedings{xing2008towardsystems,
  author    = {Xing, G and Jia, W and Du, Y and Tso, FP and Sha, M and Liu, X},
  booktitle = {IEEE International Conference on Systems, Man and Cybernetics (ICSMC)},
  note      = {date: 2008-10-06
               keywords: publications
               pubstate: published
               tppubtype: conference
               },
  title     = {Toward ubiquitous Video-based Cyber-Physical Systems},
  url       = {http://www.poscotso.com/wp-content/uploads/2016/03/xing2008towards.pdf},
  year      = {2008},
}

@mastersthesis{tso2007mphil,
  author = {Tso, Fung Po},
  title  = {3G and Internet protocols integration},
  school = {City University of Hong Kong},
  year   = {2007},
}

@inproceedings{jia2007anysenseapplications,
  author    = {Jia, W and Xing, G and Tso, FP and Dai, B},
  booktitle = {ACM Conference on Embedded Networked Sensor Systems (SenSys)},
  note      = {date: 2007-11-06
               keywords: publications
               pubstate: published
               tppubtype: conference
               },
  title     = {AnySense: a Video Communication Architecture for Urban Sensing Applications},
  url       = {http://www.poscotso.com/wp-content/uploads/2016/03/anysense-sensys07.pdf},
  year      = {2007},
}

@inproceedings{tso2007videonetworks,
  author    = {Tso, FP and Zhang, L and Jia, W},
  booktitle = {ACM Conference on Embedded Networked Sensor Systems (SenSys )},
  note      = {date: 2007-06-09
               keywords: publications
               pubstate: published
               tppubtype: conference
               },
  title     = {Video Surveillance Patrol Robot System in 3G Internet and Sensor Networks},
  url       = {http://www.poscotso.com/wp-content/uploads/2016/03/p395-tso.pdf},
  year      = {2007},
}

@article{jia2006efficientconferencing,
  author  = {Jia, W and Tso, FP and Zhang, L},
  journal = {Journal of Networks},
  note    = {date: 2006-10-02
             keywords: publications
             pubstate: published
             tppubtype: article
             },
  title   = {Efficient 3G324M protocol Implementation for Low Bit Rate Multipoint Video Conferencing},
  url     = {http://www.poscotso.com/wp-content/uploads/2016/03/jnw-jia.pdf},
  year    = {2006},
  doi     = {10.4304/jnw.1.5.1-8},
  issn    = {1796-2056},
}

@inproceedings{han2006performancenetworks,
  author    = {Han, B and Tso, FP and Ling, L and Jia, W},
  booktitle = {IEEE International Conference on Mobile Adhoc and Sensor Systems (MASS)},
  note      = {date: 2006-10-09
               keywords: publications
               pubstate: published
               tppubtype: conference
               },
  title     = {Performance Evaluation of Scheduling in IEEE 802.16 Based Wireless Mesh Networks},
  url       = {http://www.poscotso.com/wp-content/uploads/2016/03/han-2006.pdf},
  year      = {2006},
  doi       = {https://doi.org/10.1109/MOBHOC.2006.278652},
}

@incollection{jia2006videoconferencing,
  author    = {Jia, Weijia and Tso, Fung Po and Zhang, Lizhuo and Du, Yang},
  title     = {Videoconferencing for 3G Wireless Networks},
  booktitle = {Handbook on Mobile Ad Hoc and Pervasive Communications},
  year      = {2006},
}