基于遗传算法的 Docker 集群调度策略

林伟伟 王泽涛

doi:10.3969/j.issn.1000-565X.2018.03.018

华南理工大学学报(自然科学版) >

2018 , Vol. 46 >Issue 3: 127 - 133

DOI: https://doi.org/10.3969/j.issn.1000-565X.2018.03.018

计算机科学与技术

基于遗传算法的 Docker 集群调度策略

展开

华南理工大学计算机科学与工程学院，广东广州 510006

林伟伟(1980-)，男，博士，教授，主要从事分布式系统、云计算和大数据研究

收稿日期: 2017-08-18

修回日期: 2018-02-08

网络出版日期: 2018-03-01

基金资助

国家自然科学基金资助项目(61772205， 61402183);广东省科技计划项目(2017A010101008， 2017B010126002， 2017A010101014， 2017B010126002， 2017B090901061);广州市科技计划项目(201607010048， 201604010040)

收起

Docker Cluster Scheduling Strategy Based on Genetic Algorithm

Expand

School of Computer Science and Engineering，South China University of Technology，Guangzhou 510006，Guangdong，China

林伟伟(1980-)，男，博士，教授，主要从事分布式系统、云计算和大数据研究

Received date: 2017-08-18

Revised date: 2018-02-08

Online published: 2018-03-01

Supported by

Supported by the National Natural Science Foundation of China(61772205， 61402183) and the Science and Technology Planning Project of Guangdong Province(2017A010101008， 2017B010126002， 2017A010101014， 2017B010126002， 2017B090901061)

Fold

摘要

Docker 集群技术因其轻量级、部署简单、高效等特点而成为构建云计算平台的新方案．为了提升传统 Docker 集群调度策略的负载均衡性能和增加多任务并发调度能力，文中提出了基于遗传算法的 Docker 集群调度策略．该策略将多个任务合并成一个调度组，并生成相应调度结果作为种群个体; 然后结合任务负载模式、节点当前负载状态及硬件性能计算集群负载均衡值并作为个体适应度; 最后利用遗传算法筛选出全局近似最优解作为调度结果．实验结果表明，文中策略与目前流行的 Docker Swarm 的 Spread 策略、权值调度策略相比，负载均衡性能和多任务调度效率均有了显著的提高．

关键词： Docker; 容器技术; 调度策略; 遗传算法; 负载均衡

本文引用格式

林伟伟王泽涛 . 基于遗传算法的 Docker 集群调度策略[J]. 华南理工大学学报(自然科学版), 2018 , 46(3) : 127 -133 . DOI: 10.3969/j.issn.1000-565X.2018.03.018

Abstract

Docker cluster technology is a new scheme for building a cloud computing platform because of its lightness，simple deployment and high efficiency． In order to improve the load balancing performance of traditional Docker cluster scheduling strategy and increase the capacity of multi task concurrent scheduling，a genetic algo

rithm based Docker cluster scheduling strategy is proposed in this paper． This strategy will merge into a multi task scheduling group，and generate the corresponding scheduling results as individuals． Then by combining the task load mode， the current node load state and hardware performance and by calculating the load balancing value as the fitness of individuals，the genetic algorithm is finally selected as the global optimal solution of scheduling results．

Experimental results show that compared with the popular Docker swarm spread strategy and weight scheduling strategy， the proposed strategy has significantly improved load balancing performance and multitask scheduling efficiency．

Key words： Docker; container technology; scheduling strategy; genetic algorithm; load balancing

参考文献

Options

文章导航

模态框（Modal）标题

摘要

本文引用格式

Abstract

参考文献