| 本文已被：浏览 5335次 下载 7074次 |
码上扫一扫！ |
|
|
| 对象存储系统中自适应的元数据负载均衡机制 |
|
陈涛1,2,3,4, 肖侬1, 刘芳1
|
|
1.国防科学技术大学 计算机学院,湖南 长沙 410073;2.事医学科学院 放射与辐射医学研究所,北京 100850;3.蛋白质组学国家重点实验室(军事医学科学院),北京 102206;4.北京蛋白质组研究中心,北京 102206
|
|
| 摘要: |
| 面向对象的存储系统在研究、工程以及服务领域均得到了广泛的应用.在面向对象的存储系统中,元数据的负载均衡对于提高整个系统的I/O性能具有重要的作用.现有的元数据负载均衡策略不能动态地平衡元数据的访问负载,而且自适应性以及容错特性有待提高.提出了一种自适应的分布式元数据负载均衡机制(adaptabledistributed load balancing of metadata,简称ADMLB),包含基本的负载均衡算法和分布式的增量负载均衡算法.采用基本的负载均衡算法按照服务器的性能公平地分布负载,使用分布式的负载均衡算法定时地调整负载的分布.ADMLB采取分布式的方法均衡地在元数据服务器之间分布负载,根据负载的变化自适应地进行调整,具有很好的容错特性,而且用户可以高效地定位元数据服务器. |
| 关键词: 面向对象的存储系统 元数据 负载均衡 自适应 分布式 |
| DOI：10.3724/SP.J.1001.2013.04177 |
| 分类号: |
| 基金项目:国家自然科学基金(61025009,60903040,61070198,61170288);国家高技术研究发展计划(863)(2011AA010500) |
|
| Adaptive Metadata Load Balancing for Object Storage Systems |
|
CHEN Tao1,2,3,4, XIAO Nong1, LIU Fang1
|
|
1.College of Computer, National University of Defense Technology, Changsha 410073, China;2.Beijing Institute of Radiation Medicine, Beijing 100850, China;3.State Key Laboratory of Proteomics, Beijing 102206, China;4.Beijing Proteome Research Center, Beijing 102206, China
|
| Abstract: |
| Object-Based storage is a good choice for large scale storage systems. Load balancing of metadata is important to improve the performance of I/O. The existing load balancing schemas cannot evenly distribute the accesses of metadata in a dynamic way. Moreover, the adaptability and fault-tolerance ability need to be improved. This paper presents an adaptable distributed load balancing of metadata (ADMLB) which is composed of basic load balancing algorithm (BBLA) and distributed incremental load balancing algorithm (IBLA). Specially, ADMLB first uses BBLA to distribute metadata loads according to the performances of the metadata servers and then uses IBLA to incrementally reorganize loads on each metadata server. ADMLB can evenly distribute loads between metadata servers and adapts well to the changes of loads. It also has good fault-tolerance ability, and locates metadata servers very quickly. |
| Key words: object storage system metadata load balancing adaptive distributed |
