Deep Reinforcement Learning Based Delay-Sensitive Task Scheduling and Resource Management Algorithm for Multi-User Mobile-Edge Computing Systems

Meng, Hao; Chao, Daichong; Huo, Ru; Guo, Qianying; Li, Xiaowei; Huang, Tao

doi:10.1145/3325730.3325731

Cited by 8 publications

(6 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In [37], an intelligent offloading system for vehicular edge computing by leveraging DRL was constructed, in which both communication and computation states were modelled by the finite Markov chains. In [38], the authors investigated the problem of delay sensitive task scheduling and resource management on the server side in multiuser MEC scenario, where a new online algorithm based on DRL was devised to reduce average slowdown and average timeout period of tasks in the queue. In [39], the computing aware scheduling strategy in MEC was proposed, in which a support vector machine based multiclass classifier was adopted.…”

Section: Drl-based Methodsmentioning

confidence: 99%

DRL-Based Edge Computing Model to Offload the FIFA World Cup Traffic

Che

2020

Mobile Information Systems

View full text Add to dashboard Cite

In recent years, the volume of global video traffic has been increasing rapidly and it is considerably significant to offload the traffic during the process of video transmission and improve the experience of users. In this paper, we propose a novel traffic offloading strategy to provide a feasible and efficient reference for the following 2022 FIFA World Cup held in Qatar. At first, we present the system framework based on the Mobile Edge Computing (MEC) paradigm, which supports transferring the FIFA World Cup traffic to the mobile edge servers. Then, the Deep Reinforcement Learning (DRL) is used to provide the traffic scheduling method and minimize the scheduling time of application programs. Meanwhile, the task scheduling operation is regarded as the process of Markov decision, and the proximal policy optimization method is used to train the Deep Neural Network in the DRL. For the proposed traffic offloading strategy, we do the simulation based on two real datasets, and the experimental results show that it has smaller scheduling time, higher bandwidth utilization, and better experience of user than two baselines.

show abstract

Section: Drl-based Methodsmentioning

confidence: 99%

DRL-Based Edge Computing Model to Offload the FIFA World Cup Traffic

Che

2020

Mobile Information Systems

View full text Add to dashboard Cite

show abstract

“…The four benchmarks are: 1) Random, which selects ONU requests randomly, 2) the shortest request first algorithm (SRF), which serves ONU requests in the ascending order of their duration [30], 3) the resource wrapper Packer algorithm [31], which assigns resource according to the order of alignment between resource requirements and resource availability, 4) the synthesis Tetris algorithm, which balances the advantages of taking short-term request and resource packaging [32].…”

Section: A Simulation Setupmentioning

confidence: 99%

“…Extensive numerical simulations are conducted to evaluate the performance of our proposed DRL-based DSA scheme. Note that the bassline heuristics used in the simulation part, such as Tetris, SJF, Packer and Random [30][31][32], generally use fixed policy to schedule the serving order of ONU requests. When network state changes, these fixed policies cannot adapt to the changes of the network state.…”

Section: Introductionmentioning

confidence: 99%

Dynamic Subcarrier Assignment in OFDMA-PONs Based on Deep Reinforcement Learning

Zhu

Chen

et al. 2022

IEEE Photonics J.

View full text Add to dashboard Cite

Orthogonal Frequency Division Multiplexing Access Passive Optical Networks (OFDMA-PONs), a solution for the next-generation optical access network, allows multiple optical network units (ONUs) to dynamically share subcarriers (SCs) to support efficient bandwidth allocation. In uplink transmission, multiple ONUs can share orthogonal low bit rate SCs to transmit data at different time slots (TSs) during the transmission cycle. In this paper, the dynamic subcarrier allocation (DSA) scheme based on deep reinforcement learning (DRL) is proposed for various ONU bandwidth requests. The novel scheme jointly allocates time slots, subcarriers, and modulation formats in a dynamic and flexible manner. The ONU can save transmit power by using a lower order modulation format while meeting the delay requirement. The simulation part demonstrates how the proposed DRL-based DSA scheme can be adapted to various situations, including 1) variation in the size of ONU bandwidth requests, and 2) variation in the weight of different indicators. The extensive simulation results show that, for the first time, the proposed DRL-based DSA algorithm achieves optimal traffic latency with substantial power saving, compared with the traditional two-dimensional DSA algorithms.

show abstract

“…Scheduling and resource management algorithm for multi-user mobile-edge computing systems Meng et al (2019) The problem of delay-sensitive task scheduling and resource (e.g., CPU, memory) management on the server side in multi-user MEC scenario…”

Section: Algorithm Idea Advantagementioning

confidence: 99%

“… Wei et al (2018) proposed an intelligent QoS aware job scheduling framework based on Deep Q-Learning algorithm, which can effectively reduce the average response time of jobs under varying loads and improve user satisfaction. Meng et al (2019) designed an adaptive online scheduling algorithm by combining reinforcement learning with DNN, which significantly improved the scheduling efficiency of server-side task queues. Ran et al (2019) used the Deep Determining Policy Gradient (DDPG) algorithm to find the optimal task assignment scheme meeting the requirements of the Service Level Agreement (SLA).…”

Section: Related Workmentioning

confidence: 99%

A resource scheduling method for reliable and trusted distributed composite services in cloud environment based on deep reinforcement learning

Lei

Duan

et al. 2022

Front. Genet.

View full text Add to dashboard Cite

With the vigorous development of Internet technology, applications are increasingly migrating to the cloud. Cloud, a distributed network environment, has been widely extended to many fields such as digital finance, supply chain management, and biomedicine. In order to meet the needs of the rapid development of the modern biomedical industry, the biological cloud platform is an inevitable choice for the integration and analysis of medical information. It improves the work efficiency of the biological information system and also realizes reliable and credible intelligent processing of biological resources. Cloud services in bioinformatics are mainly for the processing of biological data, such as the analysis and processing of genes, the testing and detection of human tissues and organs, and the storage and transportation of vaccines. Biomedical companies form a data chain on the cloud, and they provide services and transfer data to each other to create composite services. Therefore, our motivation is to improve process efficiency of biological cloud services. Users’ business requirements have become complicated and diversified, which puts forward higher requirements for service scheduling strategies in cloud computing platforms. In addition, deep reinforcement learning shows strong perception and continuous decision-making capabilities in automatic control problems, which provides a new idea and method for solving the service scheduling and resource allocation problems in the cloud computing field. Therefore, this paper designs a composite service scheduling model under the containers instance mode which hybrids reservation and on-demand. The containers in the cluster are divided into two instance modes: reservation and on-demand. A composite service is described as a three-level structure: a composite service consists of multiple services, and a service consists of multiple service instances, where the service instance is the minimum scheduling unit. In addition, an improved Deep Q-Network (DQN) algorithm is proposed and applied to the scheduling algorithm of composite services. The experimental results show that applying our improved DQN algorithm to the composite services scheduling problem in the container cloud environment can effectively reduce the completion time of the composite services. Meanwhile, the method improves Quality of Service (QoS) and resource utilization in the container cloud environment.

show abstract

Deep Reinforcement Learning Based Delay-Sensitive Task Scheduling and Resource Management Algorithm for Multi-User Mobile-Edge Computing Systems

Cited by 8 publications

References 14 publications

DRL-Based Edge Computing Model to Offload the FIFA World Cup Traffic

DRL-Based Edge Computing Model to Offload the FIFA World Cup Traffic

Dynamic Subcarrier Assignment in OFDMA-PONs Based on Deep Reinforcement Learning

A resource scheduling method for reliable and trusted distributed composite services in cloud environment based on deep reinforcement learning

Contact Info

Product

Resources

About