MOSIX机群文件系统研究
|
摘要
机群文件系统提供的文件访问效率对整个机群系统的性能有着至关重要的影响，MOSIX负载均衡机群系统利用抢占式进程迁移和直接文件系统访问(DFSA)机制，实现了机群节点间的负载均衡和资源的协调使用，MOSIX机群文件系统MFS通过在服务器节点进行全局唯一的缓存实现了单一节点一致性，提供了对DFSA机制的支持。但这种缓存方法导致MOSIX节点使用的传统Linux虚拟文件系统缓存策略不能很好的满足MFS缓存要求，对MFS的文件访问效率产生了影响。
本文首先分析了MOSIX中单节点使用的Linux虚拟文件系统缓存机制，对机群文件系统中的两类缓存模型：三层缓存体系结构和协作式缓存进行了详细的讨论；在对MOSIX及其机群文件系统MFS和DFSA机制进行深入研究的基础上，本文把协作缓存机制引入到MFS中，提出了MOSIX全局协作缓存(MGCC)模型，改进了MOSIX的缓存策略。MGCC模型的设计保证了与MOSIX原有的内存引导算法和DFSA机制不产生冲突，从而进一步保证了MGCC的有效性。文章最后对该模型的实现进行了探讨。
The efficiency of file accessing in cluster file system plays an important role in the performance of the cluster system. MOS1X load-balancing cluster system is an enhancement of Linux. Through the preemptive process migration it can balance loads variatins in the resoures usage among the nodes in the cluster. Direct File System Access, a provision that allowing a migrated process to directly access files in the current loaction, is supported by MOSIX through the MOSIX File System(MFS). To make DFSA work, A single-node consistency is required, and MFS can support the requirement by caching the only block data on the file server in the cluster system, but the traditional Linux cache mechanism on the file server can't provide enough buffer caches to both the local and the remote client processes. At the same time , there maybe have some nodes in the cluster with low load and their buffer caches aren't used efficiently. On the analysis of Linux virtual file system, cache models in cluster file system, MFS and the mechanism of DFSA, the Paper import the provesion of Cooperate Caching to MFS and put forward the mechanism of MOSIX Global Cooperate Caching(MGCC), which can make all the buffer caches in the cluster cooperate when the cluster is working and make these buffer caches utilize fully, so the overall performance of the MFS can be improved. And the paper also describe the MGCC can work with DFSA and memory ushing algorithm harmoniously.
本文首先分析了MOSIX中单节点使用的Linux虚拟文件系统缓存机制，对机群文件系统中的两类缓存模型：三层缓存体系结构和协作式缓存进行了详细的讨论；在对MOSIX及其机群文件系统MFS和DFSA机制进行深入研究的基础上，本文把协作缓存机制引入到MFS中，提出了MOSIX全局协作缓存(MGCC)模型，改进了MOSIX的缓存策略。MGCC模型的设计保证了与MOSIX原有的内存引导算法和DFSA机制不产生冲突，从而进一步保证了MGCC的有效性。文章最后对该模型的实现进行了探讨。
The efficiency of file accessing in cluster file system plays an important role in the performance of the cluster system. MOS1X load-balancing cluster system is an enhancement of Linux. Through the preemptive process migration it can balance loads variatins in the resoures usage among the nodes in the cluster. Direct File System Access, a provision that allowing a migrated process to directly access files in the current loaction, is supported by MOSIX through the MOSIX File System(MFS). To make DFSA work, A single-node consistency is required, and MFS can support the requirement by caching the only block data on the file server in the cluster system, but the traditional Linux cache mechanism on the file server can't provide enough buffer caches to both the local and the remote client processes. At the same time , there maybe have some nodes in the cluster with low load and their buffer caches aren't used efficiently. On the analysis of Linux virtual file system, cache models in cluster file system, MFS and the mechanism of DFSA, the Paper import the provesion of Cooperate Caching to MFS and put forward the mechanism of MOSIX Global Cooperate Caching(MGCC), which can make all the buffer caches in the cluster cooperate when the cluster is working and make these buffer caches utilize fully, so the overall performance of the MFS can be improved. And the paper also describe the MGCC can work with DFSA and memory ushing algorithm harmoniously.
引文
[1] Amnon Barak, Oren La'adan, Amnon Shiloh. Scalable Cluster Computing with MOSIX for LINUX. Proc. 5-th Annual Linux Expo. Atalnta, GA, 1999. pp.95-100
[2] Lior Amar, Amnon Barak and Amnon Shiloh. The MOSIX Direct File System Access Method for Supporting Scalable Cluster File Systems http://www.mosix.org 2003
[3] Michael Donald Dahelin. Serverless Network File Systems. dissertation for the degree of Doctor Berkeley University of California
[4] M.D. Dahlin, et al. Cooperative Caching: Using Remote Client Memory to Improve File System Performance. In Proceedings of the 1st Symposium on Operating Systems Design and Implementation, Nov 1994.
[5] A.Barak, A.Braverman Memory ushering in a scalable computing cluster. Journal of Microprocessors and Microsystems, 22(3-4), Aug. 1998
[6] 许广斌．基于linux的集群系统．http://www-900.ibm.com/DeveloperWorks/ 2001，7
[7] A.Barak, S.Guday and R.Wheeler. The MOSIX Distributed Operating System, Load Balancing for UNIX. Lecture Notes in Computer Science, Vol. 672, Springer-Verlag, 1993
[8] RFC References for NFS version 4. http://www.nfsv4.org/
[9] [美]扎达克(Zadok，E．)著 邱仲潘等译．Linux网络文件系统管理指南 http://ssreader.lib.xidian.edu.cn 2001
[10] Eliezer Levy and Abraham Silberschatz. Distributed File Systems: Concepts and Examples. ACM Computing Surveys, Vol. 22, No 4, Dec. 1990. pp. 321-374,
[11] Ammon Barak, Avner Braverman Ilia Gildemame. Performance of PVM with the MOSIX preemptive process migration scheme. 1996
[12] L.Amar, A.Barak and A.Shiloh. The MOSIX Parallel I/O System for Scalable I/O Performance. Proc. 14-th IASTED International Conference on Parallel and Distributed Computing and Systems (PDCS 2002), pp. 495-500, Cambridge, MA, Nov. 2002
[13] 冯军．机群文件系统性能优化中的关键问题研究．北京．中国科学院计算技术研究所硕士学位论文 2001
[14] 徐志伟．机群文件系统性能与正确性研究．北京．中国科学院计算技术
研究所硕士学位论文 2002
[15] 史小冬、孟丹．祝明发 COSMOS：一种可扩展单—映象机群文件系统．南京大学学报(自然科学版)．2001，10
[16] 夏煜．Linux操作系统的文件系统研究．西安．西北工业大学硕士学位论文 2000
[17] 毛德操、胡希明．LINUX内核源代码情景分析．杭州：浙江大学出版社，2001
[18] [美]博伟特(Bovet，D．P．)等编著 陈莉君等译．深入理解Linux内核，2001
[19] 黄翊、蒋江、张民选．MOSIX 进程迁移机制研究．计算机工程 2002，8 8(28)：117-119
[20] [美]布亚(Buyya，R．)著 郑纬民等译．高性能集群计算(卷一)．2001．6
[21] 黄仲伟、李莲治．一种分布式文件系统及多缓存数据一致性策略．小型微型计算机系统．16(6)．1995
[22] 高兰．集群系统UDP协议单拷贝技术的研究．华中科技大学硕士学位论文 2002
[23] MOSIX．http://www.mosix.org．
[24] 何军、田范江、王鼎兴．一种机群网络文件系统的合作高速缓存技术．计算机学报．20(10)．1997，10
[25] 赵蔚．Ext2文件系统的硬盘布局．http://www-900.ibm.com/ 2002．6
[2] Lior Amar, Amnon Barak and Amnon Shiloh. The MOSIX Direct File System Access Method for Supporting Scalable Cluster File Systems http://www.mosix.org 2003
[3] Michael Donald Dahelin. Serverless Network File Systems. dissertation for the degree of Doctor Berkeley University of California
[4] M.D. Dahlin, et al. Cooperative Caching: Using Remote Client Memory to Improve File System Performance. In Proceedings of the 1st Symposium on Operating Systems Design and Implementation, Nov 1994.
[5] A.Barak, A.Braverman Memory ushering in a scalable computing cluster. Journal of Microprocessors and Microsystems, 22(3-4), Aug. 1998
[6] 许广斌．基于linux的集群系统．http://www-900.ibm.com/DeveloperWorks/ 2001，7
[7] A.Barak, S.Guday and R.Wheeler. The MOSIX Distributed Operating System, Load Balancing for UNIX. Lecture Notes in Computer Science, Vol. 672, Springer-Verlag, 1993
[8] RFC References for NFS version 4. http://www.nfsv4.org/
[9] [美]扎达克(Zadok，E．)著 邱仲潘等译．Linux网络文件系统管理指南 http://ssreader.lib.xidian.edu.cn 2001
[10] Eliezer Levy and Abraham Silberschatz. Distributed File Systems: Concepts and Examples. ACM Computing Surveys, Vol. 22, No 4, Dec. 1990. pp. 321-374,
[11] Ammon Barak, Avner Braverman Ilia Gildemame. Performance of PVM with the MOSIX preemptive process migration scheme. 1996
[12] L.Amar, A.Barak and A.Shiloh. The MOSIX Parallel I/O System for Scalable I/O Performance. Proc. 14-th IASTED International Conference on Parallel and Distributed Computing and Systems (PDCS 2002), pp. 495-500, Cambridge, MA, Nov. 2002
[13] 冯军．机群文件系统性能优化中的关键问题研究．北京．中国科学院计算技术研究所硕士学位论文 2001
[14] 徐志伟．机群文件系统性能与正确性研究．北京．中国科学院计算技术
研究所硕士学位论文 2002
[15] 史小冬、孟丹．祝明发 COSMOS：一种可扩展单—映象机群文件系统．南京大学学报(自然科学版)．2001，10
[16] 夏煜．Linux操作系统的文件系统研究．西安．西北工业大学硕士学位论文 2000
[17] 毛德操、胡希明．LINUX内核源代码情景分析．杭州：浙江大学出版社，2001
[18] [美]博伟特(Bovet，D．P．)等编著 陈莉君等译．深入理解Linux内核，2001
[19] 黄翊、蒋江、张民选．MOSIX 进程迁移机制研究．计算机工程 2002，8 8(28)：117-119
[20] [美]布亚(Buyya，R．)著 郑纬民等译．高性能集群计算(卷一)．2001．6
[21] 黄仲伟、李莲治．一种分布式文件系统及多缓存数据一致性策略．小型微型计算机系统．16(6)．1995
[22] 高兰．集群系统UDP协议单拷贝技术的研究．华中科技大学硕士学位论文 2002
[23] MOSIX．http://www.mosix.org．
[24] 何军、田范江、王鼎兴．一种机群网络文件系统的合作高速缓存技术．计算机学报．20(10)．1997，10
[25] 赵蔚．Ext2文件系统的硬盘布局．http://www-900.ibm.com/ 2002．6