计算机应用 ›› 2018, Vol. 38 ›› Issue (2): 557-562.DOI: 10.11772/j.issn.1001-9081.2017081943

• 先进计算 • 上一篇    下一篇

基于Docker的云资源弹性调度策略

彭丽苹, 吕晓丹, 蒋朝惠, 彭成辉   

  1. 贵州大学 计算机科学与技术学院, 贵阳 550025
  • 收稿日期:2017-08-09 修回日期:2017-09-23 出版日期:2018-02-10 发布日期:2018-02-10
  • 通讯作者: 吕晓丹
  • 作者简介:彭丽苹(1990-),女,湖南郴州人,硕士研究生,主要研究方向:云计算、大数据;吕晓丹(1970-),男,上海人,副教授,硕士,主要研究方向:云计算、算法设计、数据分析;蒋朝惠(1965-),男,四川广安人,教授,硕士,主要研究方向:云计算、信息安全、数据库、软件工程;彭成辉(1990-),男,湖南娄底人,硕士研究生,主要研究方向:云计算、数据挖掘。
  • 基金资助:
    贵州省基础研究重大项目(黔科合JZ字[2014]2001-21)。

Elastic scheduling strategy for cloud resource based on Docker

PENG Liping, LYU Xiaodan, JIANG Chaohui, PENG Chenghui   

  1. College of Computer Science and Technology, Guizhou University, Guiyang Guizhou 550025, China
  • Received:2017-08-09 Revised:2017-09-23 Online:2018-02-10 Published:2018-02-10
  • Supported by:
    This work is partially supported by Major Program for Basic Research of Guizhou Province (黔科合JZ字[2014]2001-21).

摘要: 针对云资源弹性调度问题,结合Ceph数据存储的特点,提出一种基于Docker容器的云资源弹性调度策略。首先,指出Docker容器数据卷不能跨主机的特性给应用在线迁移带来了困难,并对Ceph集群的数据存储方法进行改进;然后,建立了一个基于节点综合负载的资源调度优化模型;最后,将Ceph集群和Docker容器的特点相结合,利用Docker Swarm实现了既考虑数据存储、又考虑集群负载的应用容器部署算法和应用在线迁移算法。实验结果表明,与一些调度策略相比,该调度策略对集群资源进行了更细粒度的划分,实现了云平台资源的弹性调度,并在保证应用性能的同时,达到了合理利用云平台资源和降低数据中心运营成本的目的。

关键词: 弹性调度, 应用迁移, 数据放置策略, Swarm编排器, Docker容器, Ceph集群

Abstract: Considering the problem of elastic scheduling for cloud resources and the characteristics of Ceph data storage, a cloud resource elastic scheduling strategy based on Docker container was proposed. First of all, it was pointed out that the Docker container data volumes are unable to work across different hosts, which brings difficulty to apply online migration, then the data storage method of Ceph cluster was improved. Furthermore, a resource scheduling optimization model based on the comprehensive load of nodes was established. Finally, by combining the characteristics of Ceph cluster and Docker container, the Docker Swarm orchestration was used to achieve container deployment and application online migration in consideration of both data storage and cluster load. The experimental results show that compared with some scheduling strategies, the proposed scheduling strategy achieves elastic scheduling of the cloud platform resources by making a more granular partitioning of the cluster resources, makes a reasonable utilization of the cloud platform resources and reduces the cost of data center operations under the premise of ensuring the application performance.

Key words: elastic scheduling, application migration, data-place policy, Docker Swarm, Docker container, Ceph cluster

中图分类号: