• 中国计算机学会会刊
  • 中国科技核心期刊
  • 中文核心期刊

计算机工程与科学

• 论文 • 上一篇    下一篇

基于骨干网的数据中心网络服务规划设计

李耀芳1,2,吴斌2,肖杰2,李玮1,刘琦1,孙莹光1   

  1. (1.天津城建大学计算中心,天津 300384;2.天津大学计算机科学与技术学院,天津 300072)
  • 收稿日期:2015-04-29 修回日期:2016-03-29 出版日期:2017-07-25 发布日期:2017-07-25
  • 基金资助:

    国家自然科学基金(61372085)

Data center network service planning
based on backbone networks

LI Yao-fang1,2,WU Bin2,XIAO Jie2,LI Wei1,LIU Qi1,SUN Ying-guang1   

  1. (1.Computing Center,Tianjin Chengjian University,Tianjin 300384;
    2.School of Computer Science and Technology,Tianjin University,Tianjin 300072,China)
     
     
  • Received:2015-04-29 Revised:2016-03-29 Online:2017-07-25 Published:2017-07-25

摘要:

数据中心是云计算的核心,而当前基于电交换器、传统多级交换网络、集中放置与管理的数据中心架构无法满足未来云服务对高性能数据中心在可生存性、高可用性与设计灵活性等方面的要求。以网络可生存性和最小化网络代价为目标,针对数据中心的放置、服务路由及保护进行联合优化设计。首先通过设计ILP获取最优解。该ILP集成了p-cycle、服务量备份以及快速重路由等思想,分别针对单个链路或单个服务器损坏进行快速保护。然后进一步给出一种启发式算法,该算法包含数据中心的放置及服务路由和快速保护两大步骤。ILP和启发式两种方法最终都通过广泛的仿真实验进行了验证。
 

关键词: 数据中心, 光网络, 云服务, ILP, p-cycle, 可生存性

Abstract:

Data centers are the key of cloud computing. Current data center architectures, which are based on electronic switches, conventional multi-stage switch network, and centralized deployment and management, cannot meet the requirements of survivability, high-availability and design flexibility of high-performance data centers for future cloud services. We optimize data center placement with service routing and protection to achieve survivability and cost minimization. An integer linear program (ILP) is first formulated to achieve optimal design. It integrates preconfigured protection cycle (p-cycle) for fast protection against a single link failure, and a data center replicas and fast service rerouting against a service failure. We then propose a two-step heuristic algorithm for large-size network scenarios. The first step solves data center placement and service routing problem in the failure-free scenario, and the second step takes fast service protection into account. The proposed design is validated by extensive simulations.
 

Key words: data center, optical network, cloud service, ILP, p-cycle, survivability