This page has only limited features, please log in for full access.
In recent years, distributed machine learning in WANs (DML-WANs), i.e., collaboratively training a high-quality ML model cross geo-distributed micro-clouds or edge devices, has attracted attention and been widely applied. Compared with cloud-centric training, DML-WANs avoids the high cost of transferring large amounts of raw data to a central cloud and privacy concerns. However, performing DML-WANs still faces challenges. Model synchronization, an essential step of DML-WANs, is accompanied by a lot of model communication cross limited-bandwidth WANs, which generates high communication overhead. Moreover, the parameter server system, which has been widely used, performs model synchronization in a centralized manner, resulting in serious communication in-cast problem. Such communication in-cast further raises the communication overhead, leading to the low efficiency of DML-WANs. To alleviate the communication in-cast, existing researches attempt to build tree-based communication overlays over the parameter server and workers. However, we identify that these approaches can not adapt to the dynamic and heterogeneous network of DML-WANs, resulting in insufficient improvements. This paper proposes TSEngine, an adaptive communication scheduler for efficient communication overlay of the parameter server system in DML-WANs. Its core idea is to dynamically schedule the communication logic over the parameter server and workers based on the active network perception. Specifically, we propose novel communication scheduling protocols for model distribution and model aggregation, respectively. We have implemented TSEngine in a mainstream parameter server system and verified its effectiveness in DML-WANs testbeds.
Huaman Zhou; Weibo Cai; Zonghang Li; Hongfang Yu; Ling Liu; Long Luo; Gang Sun. TSEngine: Enable Efficient Communication Overlay in Distributed Machine Learning in WANs. IEEE Transactions on Network and Service Management 2021, PP, 1 -1.
AMA StyleHuaman Zhou, Weibo Cai, Zonghang Li, Hongfang Yu, Ling Liu, Long Luo, Gang Sun. TSEngine: Enable Efficient Communication Overlay in Distributed Machine Learning in WANs. IEEE Transactions on Network and Service Management. 2021; PP (99):1-1.
Chicago/Turabian StyleHuaman Zhou; Weibo Cai; Zonghang Li; Hongfang Yu; Ling Liu; Long Luo; Gang Sun. 2021. "TSEngine: Enable Efficient Communication Overlay in Distributed Machine Learning in WANs." IEEE Transactions on Network and Service Management PP, no. 99: 1-1.
As people pay more attention to environment protection, the number of electric vehicles (EVs) is gradually increasing. Energy trading management for EVs is becoming a challenge. However, existing research has not considered the problem of information sharing between energy traders and issues surrounding the protection of user privacy. Therefore, in this paper, we propose a vehicle-to-vehicle (V2V) and vehicle-to-grid (V2G) electricity trading architecture based on blockchain. All energy transactions of EVs can be recorded on the blockchain ledger to ensure privacy and smart contracts work as agents for pricing and optimal energy allocation. Furthermore, we introduce a two-way auction mechanism based on the Bayesian game and design a new price adjustment strategy. Finally, we propose a bidirectional auction mechanism based on the Bayesian game approach. We use extensive simulations to evaluate the performance of our proposed algorithm. Simulation results show that the social welfare and cost performance of our algorithm can be improved by up to 102.8 and 319%, respectively.
Long Luo; Jingcui Feng; Hongfang Yu; Gang Sun. Blockchain-Enabled Two-way Auction Mechanism for Electricity Trading in Internet of Electric Vehicles. IEEE Internet of Things Journal 2021, PP, 1 -1.
AMA StyleLong Luo, Jingcui Feng, Hongfang Yu, Gang Sun. Blockchain-Enabled Two-way Auction Mechanism for Electricity Trading in Internet of Electric Vehicles. IEEE Internet of Things Journal. 2021; PP (99):1-1.
Chicago/Turabian StyleLong Luo; Jingcui Feng; Hongfang Yu; Gang Sun. 2021. "Blockchain-Enabled Two-way Auction Mechanism for Electricity Trading in Internet of Electric Vehicles." IEEE Internet of Things Journal PP, no. 99: 1-1.
Network function virtualization (NFV) has gained prominence in next-generation cloud computing, such as the fog-based radio access network, due to their ability to support better QoS in network service provision. However, most of the current service function chain (SFC) deployment researches do not consider the Security-Service-Level-Agreement (SSLA) in the deployment solution. Therefore, in this work, we introduce the SSLA into SFC deployment to defend attacks. Firstly, we formulate the SSLA guaranteed SFC deployment problem by using linear programming. Then, we propose the Maximal-security SFC deployment algorithm (MS) to maximize the security of the SFC deployment. However, the MS algorithm results in a high deployment cost. To reduce the deployment cost, we propose the Minimal-cost and SSLA-guaranteed SFC deployment algorithm (MCSG) to minimize the deployment while satisfying the SSLA. In order to reduce the blocking ratio caused by MCSG, the Minimal-cost and SSLA-guaranteed SFC deployment algorithm with feedback adjustment (MCSG-FA) is proposed. Finally, we evaluate our proposed algorithms through simulations. The simulation results show that the blocking ratio and the deployment cost of our algorithms are better than that of the existing algorithm when meeting the SSLAs.
Dongcheng Zhao; Long Luo; Hongfang Yu; Victor Chang; Rajkumar Buyya; Gang Sun. Security-SLA-guaranteed service function chain deployment in cloud-fog computing networks. Cluster Computing 2021, 1 -16.
AMA StyleDongcheng Zhao, Long Luo, Hongfang Yu, Victor Chang, Rajkumar Buyya, Gang Sun. Security-SLA-guaranteed service function chain deployment in cloud-fog computing networks. Cluster Computing. 2021; ():1-16.
Chicago/Turabian StyleDongcheng Zhao; Long Luo; Hongfang Yu; Victor Chang; Rajkumar Buyya; Gang Sun. 2021. "Security-SLA-guaranteed service function chain deployment in cloud-fog computing networks." Cluster Computing , no. : 1-16.
Distributed machine learning is a mainstream system to learn insights for analytics and intelligence services of many fronts (e.g., health, streaming and business) from their massive operational data. In such a system, multiple workers train over subsets of data and collaboratively derive a global prediction/inference model by iteratively synchronizing their local learning results, e.g., the model gradients, which in turn generates heavy and bursty traffic and results in high communication overhead in cluster networks. Such communication overhead has became the main bottleneck that limits the efficiency of training machine learning models distributedly. In this paper, our key observation is that local gradients learned by workers may have different contributions to global model convergence and executing differential transmission for different gradients can reduce the communication overhead and improve training efficiency. However, existing gradient transmission mechanisms treat all gradients the same, which may lead to long training time. Motivated by our observations, we propose Differential Gradient Transmission (DGT), a contribution-aware differential gradient transmission mechanism for efficient distributed learning, which transfers gradients with different transmission quality according to their contributions. In addition to designing a general architecture of DGT, we have proposed a novel algorithm and a novel protocol to facilitate fast model training. Experiments on a cluster with 6 GTX 1080TI GPUs and 1Gbps network show that DGT decreases the model training time by 19.4% on GoogleNet, 34.4% on AlexNet and 36.5% on VGG-11 compared to default gradient transmission on MXNET. Its acceleration is better than the other two related transmission solutions. Besides, DGT works well with different datasets (Fashion-MNIST, Cifar10), different data distributions (IID, non-IID) and different training algorithms (BSP, FedAVG).
Huaman Zhou; Zonghang Li; Qingqing Cai; Hongfang Yu; Shouxi Luo; Long Luo; Gang Sun. DGT: A contribution-aware differential gradient transmission mechanism for distributed machine learning. Future Generation Computer Systems 2021, 121, 35 -47.
AMA StyleHuaman Zhou, Zonghang Li, Qingqing Cai, Hongfang Yu, Shouxi Luo, Long Luo, Gang Sun. DGT: A contribution-aware differential gradient transmission mechanism for distributed machine learning. Future Generation Computer Systems. 2021; 121 ():35-47.
Chicago/Turabian StyleHuaman Zhou; Zonghang Li; Qingqing Cai; Hongfang Yu; Shouxi Luo; Long Luo; Gang Sun. 2021. "DGT: A contribution-aware differential gradient transmission mechanism for distributed machine learning." Future Generation Computer Systems 121, no. : 35-47.
With the rapid development of the Internet of vehicles (IoV), routing in vehicular ad hoc networks (VANETs) has become a popular research topic. Due to the features of the dynamic network structure, constraints of road topology and variable states of vehicle nodes, VANET routing protocols face many challenges, including intermittent connectivity, large delay and high communication overhead. Location-based geographic routing is the most suitable method for VANETs, and such routing performs well on paths with an appropriate vehicle density and network load. We propose an intersection-based V2X routing protocol that includes a learning routing strategy based on historical traffic flows via Q-learning and monitoring real-time network status. The hierarchical routing protocol consists of two parts: a multidimensional Q-table, which is established to select the optimal road segments for packet forwarding at intersections; and an improved greedy strategy, which is implemented to select the optimal relays on paths. The monitoring models can detect network load and adjust routing decisions in a timely manner to prevent network congestion. This method minimizes the communication overhead and latency and ensures reliable transmission of packets. We compare our algorithm with three benchmark algorithms in an extensive simulation. The results show that our algorithm outperforms the existing methods in terms of network performance, including packet delivery ratio, end-to-end delay, and communication overhead.
Long Luo; Li Sheng; Hongfang Yu; Gang Sun. Intersection-Based V2X Routing via Reinforcement Learning in Vehicular Ad Hoc Networks. IEEE Transactions on Intelligent Transportation Systems 2021, PP, 1 -14.
AMA StyleLong Luo, Li Sheng, Hongfang Yu, Gang Sun. Intersection-Based V2X Routing via Reinforcement Learning in Vehicular Ad Hoc Networks. IEEE Transactions on Intelligent Transportation Systems. 2021; PP (99):1-14.
Chicago/Turabian StyleLong Luo; Li Sheng; Hongfang Yu; Gang Sun. 2021. "Intersection-Based V2X Routing via Reinforcement Learning in Vehicular Ad Hoc Networks." IEEE Transactions on Intelligent Transportation Systems PP, no. 99: 1-14.
Blockchain is a promising emerging technology that is envisioned to play a key role in establishing secure and reliable Internet of Things (IoT) ecosystems without the involvement of any third party. Hyperledger Fabric, a permissioned blockchain system that can yield high throughput and low consensus delay, has shown its capability in enhancing security and privacy protection for delay-sensitive Internet of Things (IoT) services. The literature however has not considered the conflicting transaction problem which may substantially limit the system performance and degrade QoS for the end users. In this paper, we propose CATP-Fabric, a new blockchain system to address the conflicting transaction problem by reducing the number of potentially conflicting transactions with less overhead. First, the transactions within a block are divided into different groups to facilitate parallel transaction processing. Then, CATP-Fabric filters stale transactions and prioritizes the read-only transactions in each group to eliminate unnecessary overhead. Finally, we formulate the selection of aborting transactions in CATP-Fabric as a binary integer programming problem and develop a low-complexity optimization algorithm to minimize the number of aborted transactions. Illustrative results show that our proposed CATP-Fabric blockchain system achieves high throughput of successful transactions while maintaining a lower aborting transaction rate compared to the benchmark blockchain systems.
Xiaoqiong Xu; Xiaonan Wang; Zonghang Li; Hongfang Yu; Gang Sun; Sabita Maharjan; Yan Zhang. Mitigating Conflicting Transactions in Hyperledger Fabric-Permissioned Blockchain for Delay-Sensitive IoT Applications. IEEE Internet of Things Journal 2021, 8, 10596 -10607.
AMA StyleXiaoqiong Xu, Xiaonan Wang, Zonghang Li, Hongfang Yu, Gang Sun, Sabita Maharjan, Yan Zhang. Mitigating Conflicting Transactions in Hyperledger Fabric-Permissioned Blockchain for Delay-Sensitive IoT Applications. IEEE Internet of Things Journal. 2021; 8 (13):10596-10607.
Chicago/Turabian StyleXiaoqiong Xu; Xiaonan Wang; Zonghang Li; Hongfang Yu; Gang Sun; Sabita Maharjan; Yan Zhang. 2021. "Mitigating Conflicting Transactions in Hyperledger Fabric-Permissioned Blockchain for Delay-Sensitive IoT Applications." IEEE Internet of Things Journal 8, no. 13: 10596-10607.
Federated Learning (FL) serves privacy-preserving collaborative learning among multiple isolated parties, while retaining their privacy data locally. Cross-device and cross-silo FL have achieved great success in cross-domain applications, in which the scarce communication resource is the primary bottleneck. Driven by the need to combine heterogeneous machines from different parties to build a shared data center, we found intra-domain FL, a new type of FL in which isolated parties collaborate in the shared data center, and strong computational heterogeneity becomes the primary bottleneck. To mitigate the training inefficiency caused by stragglers, this paper proposes an efficient synchronization algorithm ESync, which allows parties to train different iterations locally under the coordination of a novel scheduler State Server. We give the boundaries of weight divergence and optimality gap of ESync, and analyze the trade-off between convergence accuracy and communication efficiency. Extensive experiments are conducted to compare ESync with SSGD, ASGD, DC-ASGD, FedAvg, FedAsync, TiFL and FedDrop under strong computational heterogeneity. Numerical results show that ESync achieves great speed up without loss of accuracy, and therefore demonstrate the effectiveness of ESync in both training efficiency and converged accuracy.
Zonghang Li; Huaman Zhou; Tianyao Zhou; Hongfang Yu; Zenglin Xu; Gang Sun. ESync: Accelerating Intra-Domain Federated Learning in Heterogeneous Data Centers. IEEE Transactions on Services Computing 2020, PP, 1 -1.
AMA StyleZonghang Li, Huaman Zhou, Tianyao Zhou, Hongfang Yu, Zenglin Xu, Gang Sun. ESync: Accelerating Intra-Domain Federated Learning in Heterogeneous Data Centers. IEEE Transactions on Services Computing. 2020; PP (99):1-1.
Chicago/Turabian StyleZonghang Li; Huaman Zhou; Tianyao Zhou; Hongfang Yu; Zenglin Xu; Gang Sun. 2020. "ESync: Accelerating Intra-Domain Federated Learning in Heterogeneous Data Centers." IEEE Transactions on Services Computing PP, no. 99: 1-1.
Network function virtualization (NFV) in 6G can use standard virtualization techniques to enable network functions via software. Resource scheduling is one of the key research areas of NFV in 6G and is mainly used to deploy service function chains (SFCs) in substrate networks. However, determining how to utilize network resources efficiently has always been a difficult problem in SFC deployment. This paper focuses on how to efficiently provision online SFC requests in NFV with 6G. We first establish a mathematical model for the problem of online SFC provisioning. Then, we propose an efficient online service function chain deployment (OSFCD) algorithm that selects the path to deploy that is close to the SFC length. Finally, we compare our proposed algorithm with three other existing algorithms by simulation experiments. The experimental results show that the OSFCD algorithm optimizes multiple performance indicators of online SFC deployment.
Gang Sun; Zhu Xu; Hongfang Yu; Victor Chang. Dynamic Network Function Provisioning to Enable Network in Box for Industrial Applications. IEEE Transactions on Industrial Informatics 2020, 17, 7155 -7164.
AMA StyleGang Sun, Zhu Xu, Hongfang Yu, Victor Chang. Dynamic Network Function Provisioning to Enable Network in Box for Industrial Applications. IEEE Transactions on Industrial Informatics. 2020; 17 (10):7155-7164.
Chicago/Turabian StyleGang Sun; Zhu Xu; Hongfang Yu; Victor Chang. 2020. "Dynamic Network Function Provisioning to Enable Network in Box for Industrial Applications." IEEE Transactions on Industrial Informatics 17, no. 10: 7155-7164.
Blockchain has been one of the most attractive technologies for many modern and even future applications. Fabric, an open-source framework to implement the permissioned enterprise-grade blockchain, is getting increasing attention from innovators. The latency performance is crucial to the Fabric blockchain in assessing its effectiveness. Many empirical studies were conducted to analyze this performance based on different hardware platforms. These experimental results are not comparable as they are highly dependent on the underlying networks. Moreover, theoretical analysis on the latency of Fabric blockchain still receives much less attention. This paper provides a novel theoretical model to calculate the transaction latency under various network configurations such as block size, block interval, etc. Subsequently, we validate the proposed latency model with experiments, and the results show that the difference between analytical and experimental results is as low as 6.1%. We also identify some performance bottlenecks and give insights from the developer’s perspective.
Xiaoqiong Xu; Gang Sun; Long Luo; Huilong Cao; Hongfang Yu; Athanasios V. Vasilakos. Latency performance modeling and analysis for hyperledger fabric blockchain network. Information Processing & Management 2020, 58, 102436 .
AMA StyleXiaoqiong Xu, Gang Sun, Long Luo, Huilong Cao, Hongfang Yu, Athanasios V. Vasilakos. Latency performance modeling and analysis for hyperledger fabric blockchain network. Information Processing & Management. 2020; 58 (1):102436.
Chicago/Turabian StyleXiaoqiong Xu; Gang Sun; Long Luo; Huilong Cao; Hongfang Yu; Athanasios V. Vasilakos. 2020. "Latency performance modeling and analysis for hyperledger fabric blockchain network." Information Processing & Management 58, no. 1: 102436.
To meet the increasing traffic demands characterized by large bandwidth and high burstiness, more traffic has been moving to inter-datacenter elastic optical networks (inter-DC EONs) for processing. The integration of two emerging paradigms, network function virtualization (NFV) and software-defined networking (SDN), enables Internet service providers (ISPs) to deploy service function chains (SFCs) from users flexibly while reducing operational and capital expenditures. This paper focuses on the problem of online SFC provisioning in inter-DC-EONs with the aim of maximizing ISP profits, where the challenge in jointly allocating IT and spectrum resources when deploying SFCs is balanced with the deployment costs of processing as many user requests as possible. We design two-phase time-efficient orchestration algorithms for online SFC requests and the strategy of SFC splitting is adopted to improve the utilization of spectrum resources on fiber links. Simulation results show that, compared with the existing algorithm, our proposed algorithms significantly shorten the deployment time, improve total profit of ISP by up to 40% and reduce the blocking probability by up to 35%.
Hongfang Yu; Zhenrong Chen; Gang Sun; Xiaojiang Du; Mohsen Guizani. Profit Maximization of Online Service Function Chain Orchestration in an Inter-Datacenter Elastic Optical Network. IEEE Transactions on Network and Service Management 2020, 18, 973 -985.
AMA StyleHongfang Yu, Zhenrong Chen, Gang Sun, Xiaojiang Du, Mohsen Guizani. Profit Maximization of Online Service Function Chain Orchestration in an Inter-Datacenter Elastic Optical Network. IEEE Transactions on Network and Service Management. 2020; 18 (1):973-985.
Chicago/Turabian StyleHongfang Yu; Zhenrong Chen; Gang Sun; Xiaojiang Du; Mohsen Guizani. 2020. "Profit Maximization of Online Service Function Chain Orchestration in an Inter-Datacenter Elastic Optical Network." IEEE Transactions on Network and Service Management 18, no. 1: 973-985.
Since its emergence, blockchain technology has received great attention because of its advantages in terms of decentralization, transparency, traceability and the ability to be tamper-proof. These advantages help blockchain become a better option for fields such as digital currency and information storage. Specifically, consortium blockchain is preferred by researchers because it provides a certain degree of access control and a supervisory mechanism. However, in real-world applications, blockchain platforms pervasively show bottlenecks such as ultrahigh energy consumption, time inefficiency, low transaction throughput, vulnerability to targeted attacks and poor fairness of user profits, which seriously influence the performance of this technology and thus hinder its development and adoption. In this paper, we try to enhance the performance of blockchain platforms by optimizing the quality of its core module, known as the consensus algorithm. To do so, we introduce proof of assets and proof of reputation to design a voting-based decentralized consensus (VDC) algorithm for consortium blockchain. Combined with the verifiable random function (VRF), VDC realizes better fairness of user profits and time efficiency with acceptable energy consumption and without sacrificing security. The simulation results show that the proposed algorithm achieves a faster consensus process and better user fairness than existing algorithms while still maintaining a negligible energy cost and adequate security.
Gang Sun; Miao Dai; Jian Sun; Hongfang Yu. Voting-Based Decentralized Consensus Design for Improving the Efficiency and Security of Consortium Blockchain. IEEE Internet of Things Journal 2020, 8, 6257 -6272.
AMA StyleGang Sun, Miao Dai, Jian Sun, Hongfang Yu. Voting-Based Decentralized Consensus Design for Improving the Efficiency and Security of Consortium Blockchain. IEEE Internet of Things Journal. 2020; 8 (8):6257-6272.
Chicago/Turabian StyleGang Sun; Miao Dai; Jian Sun; Hongfang Yu. 2020. "Voting-Based Decentralized Consensus Design for Improving the Efficiency and Security of Consortium Blockchain." IEEE Internet of Things Journal 8, no. 8: 6257-6272.
Recently, an increasing number of deep learning-based methods have been applied in recommendation. Most such methods outperform traditional methods, especially when using the natural language processing (NLP) technique with review texts. Many deep learning-based recommender systems are used to learn latent representations of reviews written by target users and reviews written for target items. They are then combined to predict the rating of the target user for the target item. However, most previously proposed review-based deep learning methods do not conform to real-world application scenarios, in which we cannot obtain the reviews of the target user for the target item (called U2I review). In real-world recommendation settings, items are always recommended to users before they have experienced them. Therefore, the review of a target user for a target item would not be available during the testing and validation process. Many methods, such as DeepCoNN and D-ATT, do not exclude the U2I review in the process of validation and testing. Therefore, the process of testing is different from real-world application scenarios, and these methods obtain substantial valuable information from the U2I review that target users write for target items. We propose a model called ADGITN and a training strategy to solve this problem. When training, the auxiliary model learns two attention distributions that the U2I reviews over user reviews and item reviews by auxiliary tasks. These two distributions are used to guide the learning of attention distributions between user reviews and item reviews of the main model. Thus, the main model could learn how to extract attention distributions between user reviews and item reviews according to the valuable information extracted from U2I reviews. During validation, only the main model works, and it could extract better attention distributions even without the help of a U2I review. Extensive experiments show the effectiveness of our model. We validate our model on the Amazon and Yelp19 datasets, and the results show that our model outperforms existing excellent models, with up to 13.8% relative improvement compared to the performance of MPCN, which is one of the best review-based deep learning models for recommendation.
Gang Sun; Yu Li; Hongfang Yu; Victor Chang. Attention distribution guided information transfer networks for recommendation in practice. Applied Soft Computing 2020, 97, 106772 .
AMA StyleGang Sun, Yu Li, Hongfang Yu, Victor Chang. Attention distribution guided information transfer networks for recommendation in practice. Applied Soft Computing. 2020; 97 ():106772.
Chicago/Turabian StyleGang Sun; Yu Li; Hongfang Yu; Victor Chang. 2020. "Attention distribution guided information transfer networks for recommendation in practice." Applied Soft Computing 97, no. : 106772.
In view of the rapid development of the Internet of Vehicles (IoV) and wireless communication technology, intelligent transportation systems play an important role in improving urban road safety, promoting the behavioral interaction between users and networks, improving the service quality and controlling the network cost. Based on the universality and real-time nature of the IoV, data collectors can cooperate with users to sense and collect relevant data within the driving range of vehicles. To achieve this goal, this paper proposes a two-tier sensing scheme around the optimization of the sensing mechanism and data processing. In the routing layer, we build a weighted graph model based on vehicle fog, and we propose a new routing strategy to maximize each vehicle's utilization. We consider that the sensing information collected by multiple vehicles could be repetitive, Cwhich would lead to many unnecessary communication flows in the network. Therefore, in the data processing layer, we consider the resource consumption in the whole fog, assign different tasks to the sensing vehicles, and filter similar information on the relay nodes of the routing paths to reduce the waste of resources in the IoV. Finally, we compare and analyze our two-tier scheme with related approaches, and the results show that our scheme has higher link utilization and lower resource consumption for a high-speed mobile network environment in the IoV.
Gang Sun; Liangjun Song; Hongfang Yu; Xiaojiang Du; Mohsen Guizani. A Two-Tier Collection and Processing Scheme for Fog-Based Mobile Crowdsensing in the Internet of Vehicles. IEEE Internet of Things Journal 2020, 8, 1971 -1984.
AMA StyleGang Sun, Liangjun Song, Hongfang Yu, Xiaojiang Du, Mohsen Guizani. A Two-Tier Collection and Processing Scheme for Fog-Based Mobile Crowdsensing in the Internet of Vehicles. IEEE Internet of Things Journal. 2020; 8 (3):1971-1984.
Chicago/Turabian StyleGang Sun; Liangjun Song; Hongfang Yu; Xiaojiang Du; Mohsen Guizani. 2020. "A Two-Tier Collection and Processing Scheme for Fog-Based Mobile Crowdsensing in the Internet of Vehicles." IEEE Internet of Things Journal 8, no. 3: 1971-1984.
Large companies operate tens of data centers (DCs) across the globe to serve their customers and store data. On the other hand, many machine learning applications need a global view of such global data to pursue high model accuracy. However, for this Geo-distributed machine learning (Geo-DML), it is infeasible to move all data together over wide-area networks (WANs) due to scarce WAN bandwidth, privacy concerns and data sovereignty laws. Therefore, most Geo-DML systems leverage geo-distributed approaches to train models, where global model synchronization is required between DCs over WAN. With the rapid increase of training data and the model sizes, it is challenging to efficiently utilize scarce and heterogeneous WAN bandwidth to synchronize models. With the advancement of optical technology, network topology becomes reconfigurable in optical WAN, which brings a new opportunity for Geo-DML training over WAN. We propose to optimize Geo-DML training with centralized joint control of the network and reconfigurable optical layers. We respectively prove the intra-job and inter-job scheduling problems are NP-hard and strongly NP-hard. For intra-job scheduling, RoWAN based on deterministic rounding algorithm, is presented to dynamically change the topology by reconfiguring the optical devices, and allocate path and rate for each flow. For inter-job scheduling, delayed SWRT is provided to schedule multiple jobs according to their priorities. The simulations in real topologies show that RoWAN reduces global model synchronization communication time of single iteration by up to 15.54%-48.2% on average in comparison with the traditional solutions. Compared to other three inter-job scheduling approaches, delayed SWRT can reduce the weighted job completion time (WJCT) by about 60%, 44.8% and 28.76%.
Ling Liu; Hongfang Yu; Gang Sun; Long Luo; Qixuan Jin; Shouxi Luo. Job scheduling for distributed machine learning in optical WAN. Future Generation Computer Systems 2020, 112, 549 -560.
AMA StyleLing Liu, Hongfang Yu, Gang Sun, Long Luo, Qixuan Jin, Shouxi Luo. Job scheduling for distributed machine learning in optical WAN. Future Generation Computer Systems. 2020; 112 ():549-560.
Chicago/Turabian StyleLing Liu; Hongfang Yu; Gang Sun; Long Luo; Qixuan Jin; Shouxi Luo. 2020. "Job scheduling for distributed machine learning in optical WAN." Future Generation Computer Systems 112, no. : 549-560.
Networking has become a well-known performance bottleneck for distributed machine learning (DML). Although lots of works have focused on accelerating the communication process of DML, they ignore the impact of the physical network on the DML performance. Concurrently, optical circuit switches (OCSes) are increasingly applied in data centers and clusters, which can fundamentally improve DML performance. It is worth noting that the non-negligible OCS reconfiguration delay makes OCS scheduling algorithms have a great impact on the upper application performance. However, existing OCS scheduling solutions are not suitable for DML jobs due to the iterative nature of DML jobs and their interleaving characteristics of communication and computation stages. Therefore, in this paper, we study the online multi-job scheduling for DML in OCS networks. Firstly, we propose heaviest-load-first (HLF), a heuristic algorithm for intra-job scheduling, which is based on the fact that the completion time of flows on the heaviest load port has a significant impact on the job completion time. Furthermore, we present Shortest Weighted Remaining Time First (SWRTF) algorithm for inter-job scheduling. In SWRTF, an available DML job is scheduled when the served job moves from communication stage to the computation stage, which significantly improves the circuit utilization. Based on large-scale simulations, we demonstrate HLF can significantly reduce the iteration communication time by up to 64.97% compared to the state-of-the-art circuit scheduler Sunflow. Besides, SWRTF can save up to 42.9%, 54.2%, 27.2% of Weighted-Job-Completion-Time (WJCT) compared to Shortest-Job-First, Baraat and Weighted-First inter-job scheduling algorithms, respectively.
Ling Liu; Hongfang Yu; Gang Sun; Huaman Zhou; Zonghang Li; Shouxi Luo. Online job scheduling for distributed machine learning in optical circuit switch networks. Knowledge-Based Systems 2020, 201-202, 106002 .
AMA StyleLing Liu, Hongfang Yu, Gang Sun, Huaman Zhou, Zonghang Li, Shouxi Luo. Online job scheduling for distributed machine learning in optical circuit switch networks. Knowledge-Based Systems. 2020; 201-202 ():106002.
Chicago/Turabian StyleLing Liu; Hongfang Yu; Gang Sun; Huaman Zhou; Zonghang Li; Shouxi Luo. 2020. "Online job scheduling for distributed machine learning in optical circuit switch networks." Knowledge-Based Systems 201-202, no. : 106002.
To introduce the opportunities brought by plug-in hybrid electric vehicles (PHEV) to the energy Internet, we propose a local vehicle-to-vehicle (V2V) energy trading architecture based on fog computing in social hotspots and model the social welfare maximization (SWM) problem to balance the interests of both charging and discharging PHEVs. Considering transaction security and privacy protection issues, we employ consortium blockchain in our designed energy trading architecture, which is different from traditional centralized power systems, to reduce the reliance on trusted third parties. Moreover, we improve the practical Byzantine fault tolerance (PBFT) algorithm and introduce it into a consensus algorithm, called delegated proof of stake (DPOS) algorithm, to design a more efficient and promising consensus algorithm, called DPOSP, which greatly reduces resource consumption and enhances consensus efficiency. To encourage PHEVs to participate in V2V energy transactions, we design an energy iterative bidirectional auction (EIDA) mechanism to resolve the SWM problem and obtain optimal charging and discharging decisions and energy pricing. Finally, we conduct extensive simulations to verify the proposed DPOSP algorithm and provide numerical results for comparison with the performance of the genetic algorithm and the Lagrange algorithm in achieving EIDA.
Gang Sun; Miao Dai; Feng Zhang; Hongfang Yu; Xiaojiang Du; Mohsen Guizani. Blockchain-Enhanced High-Confidence Energy Sharing in Internet of Electric Vehicles. IEEE Internet of Things Journal 2020, 7, 7868 -7882.
AMA StyleGang Sun, Miao Dai, Feng Zhang, Hongfang Yu, Xiaojiang Du, Mohsen Guizani. Blockchain-Enhanced High-Confidence Energy Sharing in Internet of Electric Vehicles. IEEE Internet of Things Journal. 2020; 7 (9):7868-7882.
Chicago/Turabian StyleGang Sun; Miao Dai; Feng Zhang; Hongfang Yu; Xiaojiang Du; Mohsen Guizani. 2020. "Blockchain-Enhanced High-Confidence Energy Sharing in Internet of Electric Vehicles." IEEE Internet of Things Journal 7, no. 9: 7868-7882.
Training a high-accuracy model requires trying hundreds of configurations of hyperparameters to search for the optimal configuration. It is common to launch a group of training jobs (named cojob) with different configurations at the same time and stop the jobs performing worst every stage (i.e., a certain number of iterations). Thus deep learning requires minimizing stage completion time (SCT) to accelerate the searching. To quickly complete the stages, each job in the cojob typically uses multiple GPUs to perform distributed training. The GPUs exchange data per iteration to synchronize their models through the network. However, data transfers of DL jobs compete for network bandwidth since the GPU cluster hosts a number of cojobs from various users, resulting in network congestion and consequently a large SCT for cojobs. Existing flow schedulers aimed at reducing flow/coflow/job completion time mismatch the requirement of hyperparameter searching. In this paper, we implement a system Grouper to minimize average SCT for cojobs. Grouper adopts a well-designed algorithm to permute stages of cojobs and schedules flows from different stages in the order of the permutation. The extensive testbed experiments and simulations show that Grouper outperforms advanced network designs Baraat, Sincrona, and per-flow fair share.
Pan Zhou; Hongfang Yu; Gang Sun. Grouper: Accelerating Hyperparameter Searching in Deep Learning Clusters With Network Scheduling. IEEE Transactions on Network and Service Management 2020, 17, 1879 -1895.
AMA StylePan Zhou, Hongfang Yu, Gang Sun. Grouper: Accelerating Hyperparameter Searching in Deep Learning Clusters With Network Scheduling. IEEE Transactions on Network and Service Management. 2020; 17 (3):1879-1895.
Chicago/Turabian StylePan Zhou; Hongfang Yu; Gang Sun. 2020. "Grouper: Accelerating Hyperparameter Searching in Deep Learning Clusters With Network Scheduling." IEEE Transactions on Network and Service Management 17, no. 3: 1879-1895.
Liangjun Song; Gang Sun; Hongfang Yu; Xiaojiang Du; Mohsen Guizani. FBIA: A Fog-Based Identity Authentication Scheme for Privacy Preservation in Internet of Vehicles. IEEE Transactions on Vehicular Technology 2020, 69, 5403 -5415.
AMA StyleLiangjun Song, Gang Sun, Hongfang Yu, Xiaojiang Du, Mohsen Guizani. FBIA: A Fog-Based Identity Authentication Scheme for Privacy Preservation in Internet of Vehicles. IEEE Transactions on Vehicular Technology. 2020; 69 (5):5403-5415.
Chicago/Turabian StyleLiangjun Song; Gang Sun; Hongfang Yu; Xiaojiang Du; Mohsen Guizani. 2020. "FBIA: A Fog-Based Identity Authentication Scheme for Privacy Preservation in Internet of Vehicles." IEEE Transactions on Vehicular Technology 69, no. 5: 5403-5415.
The efficient deployment of virtual network functions (VNFs) for network service provisioning is key for achieving network function virtualization (NFV); however, most existing studies address only offline or one-off deployments of service function chains (SFCs) while neglecting the dynamic (i.e., online) deployment and expansion requirements. In particular, many methods of energy/resource cost reduction are achieved by merging VNFs. However, the energy waste and device wear for large-scale collections of servers (e.g., cloud networks and data centers) caused by sporadic request updating are ignored. To solve these problems, we propose an energy-aware routing and adaptive delayed shutdown (EAR-ADS) algorithm for dynamic SFC deployment, which includes the following features. 1) Energy-aware routing (EAR): By considering a practical deployment environment, a flexible solution is developed based on reusing open servers and selecting paths with the aims of balancing energy and resources and minimizing the total cost. 2) Adaptive delayed shutdown (ADS): The delayed shutdown time of the servers can be flexibly adjusted in accordance with the usage of each device in each time slot, thus eliminating the no-load wait time of the servers and frequent on/off switching. Therefore, EAR-ADS can achieve dual energy savings by both decreasing the number of open servers and reducing the idle/switching energy consumption of these servers. Simulation results show that EAR-ADS not only minimizes the cost of energy and resources but also achieves an excellent success rate and stability. Moreover, EAR-ADS is efficient compared with an improved Markov algorithm (SAMA), reducing the average deployment time by more than a factor of 40.
Gang Sun; Run Zhou; Jian Sun; Hongfang Yu; Athanasios V. Vasilakos. Energy-Efficient Provisioning for Service Function Chains to Support Delay-Sensitive Applications in Network Function Virtualization. IEEE Internet of Things Journal 2020, 7, 6116 -6131.
AMA StyleGang Sun, Run Zhou, Jian Sun, Hongfang Yu, Athanasios V. Vasilakos. Energy-Efficient Provisioning for Service Function Chains to Support Delay-Sensitive Applications in Network Function Virtualization. IEEE Internet of Things Journal. 2020; 7 (7):6116-6131.
Chicago/Turabian StyleGang Sun; Run Zhou; Jian Sun; Hongfang Yu; Athanasios V. Vasilakos. 2020. "Energy-Efficient Provisioning for Service Function Chains to Support Delay-Sensitive Applications in Network Function Virtualization." IEEE Internet of Things Journal 7, no. 7: 6116-6131.
Victor Chang; Gang Sun; Gary Wills. Special issue on fog/edge computing in Enterprise Multimedia Security [SI 1138T]. Multimedia Tools and Applications 2020, 79, 10699 -10700.
AMA StyleVictor Chang, Gang Sun, Gary Wills. Special issue on fog/edge computing in Enterprise Multimedia Security [SI 1138T]. Multimedia Tools and Applications. 2020; 79 (15-16):10699-10700.
Chicago/Turabian StyleVictor Chang; Gang Sun; Gary Wills. 2020. "Special issue on fog/edge computing in Enterprise Multimedia Security [SI 1138T]." Multimedia Tools and Applications 79, no. 15-16: 10699-10700.