Study on RDMA-based high performance RPC in multi-core systems

被引：0
|
作者：
Li B. [1 ,2 ,3 ]
Meng D. [1 ,2 ]
Huo Z. [1 ,2 ]
机构：
[1] National Research Center for Intelligent Computing Systems, Institute of Computing Technology, Chinese Academy of Sciences
[2] Key Laboratory of Computer System and Architecture, Chinese Academy of Sciences
[3] Graduate University of Chinese Academy of Sciences
来源：
Gaojishu Tongxin/Chinese High Technology Letters | 2011年 / 21卷 / 07期
关键词：
Multi-core; Network file system (NFS); Remote direct memory access (RDMA); Remote procedure call (RPC);
D O I：
10.3772/j.issn.1002-0470.2011.07.003
中图分类号：
学科分类号：
摘要：
In consideration of the performance and scalability problems of current designs of remote procedure call (RPC) over remote direct memory access (RDMA) in multi-core systems, this paper proposes a novel design of high performance RPC based on RDMA Write and provides support to the network file system (NFS). The paper mainly focuses on the security concerns and the RDMA memory management mechanism of the design. In terms of the utilization of the RDMA memory, this paper proposes the passive reclaim strategy and the active reclaim strategy. The evaluations show that the new design proposed in this paper decreases the RPC latency by 13%-20%. The design can highly improve the RPC performance and scalability in multi-core systems.
引用
收藏
页码：681 / 686
页数：5
相关论文
共 8 条
  • [1] Sandberg R., The sun network filesystem: Design, implementation, and experience, Proceedings of the Distributed Computing Systems: Concepts and Structures, pp. 300-316, (1992)
  • [2] Ko M., Chadalapaka M., Hufferd J., Et al., iSCSI Extensions for RDMA, (2007)
  • [3] Magoutis K., Addetia S., Fedorova A., Et al., Making the most out of direct-access network attached storage, Proceedings of 2nd USENIX Conference on File and Storage Technologies, pp. 189-202, (2003)
  • [4] Goglin B., Prylli L., Performance analysis of remote file system access over a high-speed local network, Proceedings of the Workshop on Communication Architecture for Clusters, in Conjunction with International Parallel and Distributed Processing Symposium'04, pp. 185-192, (2004)
  • [5] Callaghan B., Lingutla-Raj T., Chiu A., Et al., NFS over RDMA, Proceedings of the ACM SIGCOMM Workshop on Network-I/O Convergence: Experience, Lessons, Implications, pp. 196-208, (2003)
  • [6] Noronha R., Chai L., Talpey T., Et al., Designing NFS with RDMA for Security, Performance and Scalability, Proceedings of the 2007 International Conference on Parallel Processing, pp. 49-56, (2007)
  • [7] Noronha R., Chai L., Shepler S., Et al., Enhancing the Performance of NFSv4 with RDMA, Proceedings of the International Workshop on Storage Network Architecture and Parallel I/Os, pp. 90-96, (2007)
  • [8] Yu W.K., Nageswara S.V.R., Wyckoff P., Et al., Performance of RDMA-capable storage protocols on wide-area network, Proceedings of the 2008 Petascale Data Storage Workshop, pp. 1-5, (2008)
1