冗余服务中容错算法的研究与设计
|
摘要
随着计算机网络和通信技术的发展，分布计算逐渐成为计算技术的主流。当前，大规模事务处理、分布实时以及关键抗毁等应用的发展对分布式系统的可用性和性能两方面都提出了很高的要求。在这些应用的驱动下，越来越多的分布式系统采用冗余服务技术来提高可用性和性能，冗余服务容错逐渐成为分布式领域的研究热点。
本文在较深入地研究了冗余服务中的容错算法、负载平衡的基础上，通过在冗余服务中引入主动负载平衡机制，提出了一种行之有效的冗余服务容错算法；该算法一方面能够满足客户请求对可用性的要求，另一方面有效地提高了系统的性能；基于CORBA分布计算通信平台，本文还构建了该算法的对象模型，并在此基础上实现了一个原型系统，通过实例测试验证了该算法的可行性。
本文的主要工作体现在以下几个方面：
1．在冗余服务容错中引入主动负载平衡机制，实现了资源预留的负载平衡。
2．把客户请求区分为读写两类，实现了简单高效的请求并行执行模型。
3．设计了简单高效的多备份数据管理中的状态一致性维护协议。
As development of computer network and communication technology, distributed computing becomes the main computing technology. Currently, the developments of distributed applications, such as large transaction processing, real-time distributed computing and key damage-tolerant application, require more higher availability and performance in distributed system. So the redundant service is used in more and more distributed systems for improving availability and performance, and fault-tolerant algorithms are widely studied in redundant service management system.
After studying more deeply the fault tolerant algorithm and load balancing technology in redundant services, active load balancing is introduced into the redundant service system, and a novel fault-tolerant algorithm is built in this paper. Under the conditions of guaranteeing the availability of redundant service system, the performance is effectively improved. The logic object model is based on CORBA and the prototype is implemented, additionally, system test results show the feasibility of the algorithm. The main work of this paper is as follows:
a) By introducing the active idea into redundant service, the resource-saved load balancing is performed well.
b) Requests are classified into read and write types, and both of which are dealt with accordingly. The parallel run model of requests is performed very simply and effectively.
c) The consistency protocol of state replications is performed very simply and effectively.
本文在较深入地研究了冗余服务中的容错算法、负载平衡的基础上，通过在冗余服务中引入主动负载平衡机制，提出了一种行之有效的冗余服务容错算法；该算法一方面能够满足客户请求对可用性的要求，另一方面有效地提高了系统的性能；基于CORBA分布计算通信平台，本文还构建了该算法的对象模型，并在此基础上实现了一个原型系统，通过实例测试验证了该算法的可行性。
本文的主要工作体现在以下几个方面：
1．在冗余服务容错中引入主动负载平衡机制，实现了资源预留的负载平衡。
2．把客户请求区分为读写两类，实现了简单高效的请求并行执行模型。
3．设计了简单高效的多备份数据管理中的状态一致性维护协议。
As development of computer network and communication technology, distributed computing becomes the main computing technology. Currently, the developments of distributed applications, such as large transaction processing, real-time distributed computing and key damage-tolerant application, require more higher availability and performance in distributed system. So the redundant service is used in more and more distributed systems for improving availability and performance, and fault-tolerant algorithms are widely studied in redundant service management system.
After studying more deeply the fault tolerant algorithm and load balancing technology in redundant services, active load balancing is introduced into the redundant service system, and a novel fault-tolerant algorithm is built in this paper. Under the conditions of guaranteeing the availability of redundant service system, the performance is effectively improved. The logic object model is based on CORBA and the prototype is implemented, additionally, system test results show the feasibility of the algorithm. The main work of this paper is as follows:
a) By introducing the active idea into redundant service, the resource-saved load balancing is performed well.
b) Requests are classified into read and write types, and both of which are dealt with accordingly. The parallel run model of requests is performed very simply and effectively.
c) The consistency protocol of state replications is performed very simply and effectively.
引文
[1] Khanna Raman ed. Distributed computing: implementation and management strategies. Englewood Cliffs, NJ: Prentice Hall, 1994.
[2] Silvano,M. Client/Server term definition. In: Hemmendinger, D., Reilly,E.D.,eds. Encyclopaedia of Computer Science. Zurich: International Thomson Computer Publisher,1998.
[3] Bernstein P. A. Transaction processing monitors. ACM Comm. 1990, 33(11):75～86.
[4] Bell D., Grimson J. Distributed database system. MA: Addison-Wesley Publishing Company, 1992.
[5] Rachid,G.,Andre,S. Softwre-Based replication for fault tolerance. IEEE Computer, 1997,30(4): 68～74.
[6] Michael Rabinovich. Efficient replication management in distributed systems, Ph.D.These. Department of Computer Science and Engineering, 1994.
[7] Algirdas Avizienis. Toward Systematic Design of Fault-Tolerant Systems. IEEE Computer 1997 30(4) 51～58.
[8] Richard Golding, Elizabeth Borowsky. Fault-Tolerant Replication Management in Large-Scale Distributed Storage Systems. 18th IEEE Symposium on Reliable Distributed Systems, October 18 - 21, 1999.
[9] 孙英华等．多处理机容错系统中实时任务的轮转式调度算法．计算机工程与应用，2001 17 104～106．
[10] 周明辉等．基于CORBA的容错对象复制算法．计算机研究与发展，2002．3 39(3)290～294．
[11] 唐文胜，张拥军．分布式系统中基于复制的动态容错模型．计算机工程与应用，2001 23 130～132．
[12] Arun K. Somani, Nitin H. Vaidya. Understanding fault tolerance and reliability. IEEE Computer, 1997: 45～50.
[13] 黄铠，徐志伟．可扩展并行计算技术、结构与编程．机械工业出版社，2000第一版．
[14] Michael Rabinovich. An Efficient and Highly Available Read-One Write-All Protocol for Replicated Data Management. Ph.D.These, Department of Computer Science and Engineering, University of Washington. WA 98195.
[15] Pablo Galdamez et al. High availability supports in CORBA environment. IEEE Computer, 1997: 47～49.
[16] 钱方等．提高冗余服务性能的动态容错算法．软件学报，2001 12(6)：928～935．
[17] Mustaque,A.,Mostafa,H.A. performance characterization of quorum-consensus algorithms for replicated data. IEEE Transaction on Software Engineering, 1989,15(4): 492～496.
[18] Object Management Group. The Common Object Request Broker: Architecture and Specification Revision 2.3.1. 1999.10.
[19] 陈国良等．分布式动态负载平衡调度的一个通用模型．软件学报1998 9(1)：25～29．
[20] 喻占武等．一种支持分布式进程迁移的动态负载平衡征募算法的研究．小型微型计算机系统1999 20(6)321～325．
[21] 汪芸，谢俊清，沈卓炜，顾冠群．实时CORBA技术综述．东南大学学报(自然科学版)，2002 32(3)311～317．
[22] Jie wu著，高传善等译．Distributed System Design．机械工业出版社，2001第一版．
[23] LI Guo-dong, ZHANG De-fu. A Distributed Object Based Framework for Parallel ComputatiOnS．软件学报，2002 13(3) 342～353．
[24] 毛新军等．Agent技术及其标准化．计算机科学2001．28(4)1～5．
[25] 陆月明等．soflnet——一个基于移动代理的主动网络．计算机学报，2001 21(11)，1210～1214．
[26] 杨文波等．高性能分布式双工实时容错系统中的若干技术问题．小型微型
计算机系统，2001 22(2)250～253．
[27] Thomas Delica. Modeling of some plain load distribution strategies for jobs in a multicomputer system. Information Sciences, 1997: 35～43.
[28] Andrew S. Tanenbaum. Distributed Operating System.清华大学出版社 1997 第一版．
[29] 郑剑平等．CORBA及其应用问题分析．通信学报，1999 20(10)42～48
[2] Silvano,M. Client/Server term definition. In: Hemmendinger, D., Reilly,E.D.,eds. Encyclopaedia of Computer Science. Zurich: International Thomson Computer Publisher,1998.
[3] Bernstein P. A. Transaction processing monitors. ACM Comm. 1990, 33(11):75～86.
[4] Bell D., Grimson J. Distributed database system. MA: Addison-Wesley Publishing Company, 1992.
[5] Rachid,G.,Andre,S. Softwre-Based replication for fault tolerance. IEEE Computer, 1997,30(4): 68～74.
[6] Michael Rabinovich. Efficient replication management in distributed systems, Ph.D.These. Department of Computer Science and Engineering, 1994.
[7] Algirdas Avizienis. Toward Systematic Design of Fault-Tolerant Systems. IEEE Computer 1997 30(4) 51～58.
[8] Richard Golding, Elizabeth Borowsky. Fault-Tolerant Replication Management in Large-Scale Distributed Storage Systems. 18th IEEE Symposium on Reliable Distributed Systems, October 18 - 21, 1999.
[9] 孙英华等．多处理机容错系统中实时任务的轮转式调度算法．计算机工程与应用，2001 17 104～106．
[10] 周明辉等．基于CORBA的容错对象复制算法．计算机研究与发展，2002．3 39(3)290～294．
[11] 唐文胜，张拥军．分布式系统中基于复制的动态容错模型．计算机工程与应用，2001 23 130～132．
[12] Arun K. Somani, Nitin H. Vaidya. Understanding fault tolerance and reliability. IEEE Computer, 1997: 45～50.
[13] 黄铠，徐志伟．可扩展并行计算技术、结构与编程．机械工业出版社，2000第一版．
[14] Michael Rabinovich. An Efficient and Highly Available Read-One Write-All Protocol for Replicated Data Management. Ph.D.These, Department of Computer Science and Engineering, University of Washington. WA 98195.
[15] Pablo Galdamez et al. High availability supports in CORBA environment. IEEE Computer, 1997: 47～49.
[16] 钱方等．提高冗余服务性能的动态容错算法．软件学报，2001 12(6)：928～935．
[17] Mustaque,A.,Mostafa,H.A. performance characterization of quorum-consensus algorithms for replicated data. IEEE Transaction on Software Engineering, 1989,15(4): 492～496.
[18] Object Management Group. The Common Object Request Broker: Architecture and Specification Revision 2.3.1. 1999.10.
[19] 陈国良等．分布式动态负载平衡调度的一个通用模型．软件学报1998 9(1)：25～29．
[20] 喻占武等．一种支持分布式进程迁移的动态负载平衡征募算法的研究．小型微型计算机系统1999 20(6)321～325．
[21] 汪芸，谢俊清，沈卓炜，顾冠群．实时CORBA技术综述．东南大学学报(自然科学版)，2002 32(3)311～317．
[22] Jie wu著，高传善等译．Distributed System Design．机械工业出版社，2001第一版．
[23] LI Guo-dong, ZHANG De-fu. A Distributed Object Based Framework for Parallel ComputatiOnS．软件学报，2002 13(3) 342～353．
[24] 毛新军等．Agent技术及其标准化．计算机科学2001．28(4)1～5．
[25] 陆月明等．soflnet——一个基于移动代理的主动网络．计算机学报，2001 21(11)，1210～1214．
[26] 杨文波等．高性能分布式双工实时容错系统中的若干技术问题．小型微型
计算机系统，2001 22(2)250～253．
[27] Thomas Delica. Modeling of some plain load distribution strategies for jobs in a multicomputer system. Information Sciences, 1997: 35～43.
[28] Andrew S. Tanenbaum. Distributed Operating System.清华大学出版社 1997 第一版．
[29] 郑剑平等．CORBA及其应用问题分析．通信学报，1999 20(10)42～48