分布式网络并行系统在舰载指控系统中的应用研究
|
摘要
分布式系统比单机系统具有更高的性能和可靠性，并且随着互联网的普及和某些应用本身所具有的分布式特性，使其应用日益广泛。但是，对于网络分布式系统来说，当系统的任务量过重时，单机处理速度的瓶颈、网络拥塞和延迟、负载分布的不均衡，使分布式系统达不到预期的性能要求，因此进一步提高分布式系统的性能是一个急待解决的问题。
本文结合了分布式系统和并行系统的优点，将单机结点的分布式系统改造为网络并行计算结点的分布式系统，并首次应用于海军综合电子系统的重要组成部分——水面舰载指控系统中，由于网络并行计算结点的处理速度高于单机结点，从而提高了整个分布式系统的处理速度。同时，本文还深入分析了基于网络的并行计算系统程序设计方法，对本课题所研究的舰载指控系统原有的分布式体系结构按照并行化的思想进行了优化；提出了一种适合目前海军舰载指控系统作战要求的分布式网络并行体系结构；讨论了这种体系结构的可行性和性能优势。另外，本文根据前面提出的分布式网络并行舰载指控系统体系结构的特点，通过研究目前各种并行系统负载平衡算法，提出一种适合于这种网络并行分布式体系结构的负载平衡解决方案，即基于优先级的任务请求动态负载平衡，使系统的稳定性和可靠性得到了一定的提高。
本文所提出的分布式网络并行体系结构和适应于这种体系结构的负载平衡方法在舰载指控系统仿真平台上做了仿真试验，证明这种方法的应用使舰载指控系统在性能上比原有的分布式系统有所提高，从而验证了本文所提出的方法具有一定的实用价值。
Distributed system has higher performance and reliability than the single computer system, and has beening used wider and wider with the all-pervading of Internet and distributed characteristics of some applications. But for the distributed system, when the system is overloaded, the distributed system can not meet the performance requirements anticipated because of the bottleneck of the speed of monolithic processor, network congestion and delay and unbalanced partition of load, consequently to improve the performance of the distributed system ulteriorly is an imperative problem.
This thesis combines the merits of distributed system and those of parallel system, and the monolithic processor conjunction distributed system is changed into networked parallel computing conjunction distributed system, which is applied in important component of navy integrated electronic system for the first time-surface vessel command and control system . On account of the handling speed of networked parallel computing conjunction is higher than that of monolithic processor conjunction, so the handling speed of whole distributed system is improved. At the same time, this thesis has analyzed the programming method of parallel computing system based on network and optimized the original distributed architecture of surface vessel command and control system which is studied in this thesis according to parallelism , and provided a distributed network parallel architecture which is adapt to operational requirements of actual navy surface vessel command and control system, and discussed the feasibility and performance predominance of this kind of architecture. In addition, this thesis provides a kind of load balance resolution scheme which is adapt to the networked parallel distributed architecture according to the characteristics of the architecture of distributed networked parallel surface vessel command and control system and studying actual all kinds of parallel system load balance algorithms, that is to demand dynamic load balance on the basis of prior mission and to make
the stability and reliability of the system to be improved.
The distributed networked parallel architecture and load balance resolution scheme which is adapt to this kind of architecture have been simulated on the surface vessel command and control system artificial platform, and the experiment shows that the performance of the surface vessel command and control system is higher than that of the original distributed system by making use of this kind of method and proves that the method discussed in this thesis has some practical value.
本文结合了分布式系统和并行系统的优点，将单机结点的分布式系统改造为网络并行计算结点的分布式系统，并首次应用于海军综合电子系统的重要组成部分——水面舰载指控系统中，由于网络并行计算结点的处理速度高于单机结点，从而提高了整个分布式系统的处理速度。同时，本文还深入分析了基于网络的并行计算系统程序设计方法，对本课题所研究的舰载指控系统原有的分布式体系结构按照并行化的思想进行了优化；提出了一种适合目前海军舰载指控系统作战要求的分布式网络并行体系结构；讨论了这种体系结构的可行性和性能优势。另外，本文根据前面提出的分布式网络并行舰载指控系统体系结构的特点，通过研究目前各种并行系统负载平衡算法，提出一种适合于这种网络并行分布式体系结构的负载平衡解决方案，即基于优先级的任务请求动态负载平衡，使系统的稳定性和可靠性得到了一定的提高。
本文所提出的分布式网络并行体系结构和适应于这种体系结构的负载平衡方法在舰载指控系统仿真平台上做了仿真试验，证明这种方法的应用使舰载指控系统在性能上比原有的分布式系统有所提高，从而验证了本文所提出的方法具有一定的实用价值。
Distributed system has higher performance and reliability than the single computer system, and has beening used wider and wider with the all-pervading of Internet and distributed characteristics of some applications. But for the distributed system, when the system is overloaded, the distributed system can not meet the performance requirements anticipated because of the bottleneck of the speed of monolithic processor, network congestion and delay and unbalanced partition of load, consequently to improve the performance of the distributed system ulteriorly is an imperative problem.
This thesis combines the merits of distributed system and those of parallel system, and the monolithic processor conjunction distributed system is changed into networked parallel computing conjunction distributed system, which is applied in important component of navy integrated electronic system for the first time-surface vessel command and control system . On account of the handling speed of networked parallel computing conjunction is higher than that of monolithic processor conjunction, so the handling speed of whole distributed system is improved. At the same time, this thesis has analyzed the programming method of parallel computing system based on network and optimized the original distributed architecture of surface vessel command and control system which is studied in this thesis according to parallelism , and provided a distributed network parallel architecture which is adapt to operational requirements of actual navy surface vessel command and control system, and discussed the feasibility and performance predominance of this kind of architecture. In addition, this thesis provides a kind of load balance resolution scheme which is adapt to the networked parallel distributed architecture according to the characteristics of the architecture of distributed networked parallel surface vessel command and control system and studying actual all kinds of parallel system load balance algorithms, that is to demand dynamic load balance on the basis of prior mission and to make
the stability and reliability of the system to be improved.
The distributed networked parallel architecture and load balance resolution scheme which is adapt to this kind of architecture have been simulated on the surface vessel command and control system artificial platform, and the experiment shows that the performance of the surface vessel command and control system is higher than that of the original distributed system by making use of this kind of method and proves that the method discussed in this thesis has some practical value.
引文
[1] 高传善等译，分布式系统设计．机械工业出版社，2001：20-283页
[2] 周培德编著：算法设计与分析．机械工业出版社．2002：308-329页
[3] 黄铠等，可扩展并性计算技术、结构与编程．机械工业出版社．2000：429-458页
[4] 路鑫达等译，并行程序设计．机械工业出版社．2002：31-177页
[5] 都志辉编著．高性能计算并行编程技术．清华大学出版社．2001：38-81页
[6] Rajkumar Buyya著，高性能集群计算：结构与系统(第一卷)．电子工业出版社．2001：354-445
[7] Dale Rogerson著，COM技术内幕．清华大学出版社．2000：27-79页
[8] 邓昕等，多机网络接口的性能提高，96全国计算机体系结构学术研讨会，1996：1-5
[9] 熊建新等，一种分析基于消息通讯模型的并行程序中的不确定性的方法，第六届全国青年工作者会议，1996：69-74
[10] 傅强等，基于PVM的动态任务负载平衡系统的实现，97全国计算机体系结构学术会议，1997：57-62
[11] 杜术，消息传递系统MPI检查点和进程迁移技术的研究与实现，中国学位论文数据库，2001
[12] 刘岩，并行处理与计算机网络中的组合优化问题的研究，中国学位论文数据库，1996
[13] 徐松，分布式系统中的任务迁移和动态负载均衡调整，中国学位论文数据库
[14] 周泽华，消息中间件管理研究，中国学位论文数据库，2000
[15] 李琰，韩焱，对雷达视频图象的压缩处理算法实现．船舶论证参考．2002年第3期．No．3．2002．50-54页
[16] 梁剑等，CORBA技术剖析及其应用．计算机应用研究．2002年第8期．90-93页
[17] 肖龙，实时操作系统在导航显控台中的应用．船舶论证参考．2002年第3期No．3．2002．7-11页
[18] 刘从越，利用UML获取功能需求的方法．计算机应用研究．2002年第8期．88—90页
[19] 周世兵，运用UML为软件项目建模研究．计算机应用研究．2002年第8
期．85-87页
[20]王瑞金等，统一建模语言UML及其建模实例．计算机应用研究．2002年第8期．80-84页
[21]黄汝维等，Delphi中三层结构与CORBA的应用研究计算机应用研究．2002年第8期．94-96页
[22]陈莉群等，linux操作系统内核分析．人民邮电出版社．2000：33-69页
[23]梁玉柱等，COM+技术解决方案设计．机械工业出版社．2001
[24]潘爱民等，Visual C++技术内幕．清华大学出版社．2001
[25]官章全等，Visual C++6.0类库大全．电子工业出版社．1999
[26]吴定一等，异步传递模式的理论与应用．科学出版社
[27]施振川等，UNIX网路编程(第2版)第1卷：套接口API和X./Open传输接口API．清华大学出版社．2001．9
[28]王书洪等，Windows高级编程指南．清华大学出版社．1999
[29]刘宗田等，C++编程思想．机械工业出版社．2001
[30]胡谷雨等，TCP/IP详解卷3．机械工业出版社．2001
[31]陈逸等，DCOM入门．中国电力出版社．2001
[32]潘爱民等，COM本质论．中国电力出版社．2001：33-64页
[33]L.Duchien and L.Seinturier ,OBSERVATION OF DISTRIBUTED COMUTATIONS: A REFLECTIVE APPROACH FOR CORBA,International Journal of Parallel and Distributed Systems and Networks,Vol.4.No. 1,2001 P17-25
[34]Gelff Coulson,Gordon S.Blair,Michael Clarke,Nikos Parlabvantzas ,The design of a configurable and reconfigurable middleware platform, Distributed Computing,(2002) 15:109-126
[35]熊贵喜等，计算机网络(第3版)．清华大学出版社．2000：22-138页
[36]林生，计算机通信网原理．西安电子科技大学出版社．1997
[37]陈向群等，现代操作系统．机械工业出版社．2001
[38]孔祥营等，嵌入式操作系统VxWorks及其开发环境Tornado．中国电力出版社．2002
[39](美)Behrouz Forouzam等著，潘仡等译，数据通信与网络．机械工业出版社．2000
[40]The Globus Project, www.globus.org
[41]K. Czajkowski, I. Foster, C. Kesselman, S. Martin, W. Smith, and S. Tuecke. A resource management architecture for metacomputing systems. Technical report, Mathematics and Computer Science Division, Argonne National Laboratory, Argonne, Ⅲ., 1997. Submitted.
[42] S. Fitzgerald, I. Foster, C. Kesselman, G. von Laszewski, W. Smith, and S. Tuecke. A directory service for conguring high-performance distributed computations. In Proc. 6th IEEE Symp.on High Performance Distributed Computing, pages 365-{375. IEEE Computer Society Press, 1997.
[43] I. Foster, J. Geisler, and S. Tuecke. MPI on the I-WAY: A wide-area, multimethod implementa-tion of the Message Passing Interface. In Proceedings of the 1996 MPI Developers Conference, pages 10-17. IEEE Computer Society Press, 1996.
[44] I. Foster and C. Kesselman, editors. Computational Grids: The Future of High-Performance Distributed Computing. Morgan Kaufmann Publishers, 1998.
[45] W. Gropp and E. Lusk. An abstract device denition to support the implementation of a high-level point-to-point message-passing interface. Preprint MCS-P342-1193, Mathematics and Computer Science Division, Argonne National Laboratory, Argonne, Ⅲ. 1994.
[46] W. Gropp and E, Lusk. MPICH working note: Creating a new MPICH device using the channel interface. Technical Report ANL/MCS-TM-213, Mathematics and Computer Science Division, Argonne National Laboratory, Argonne, Ⅲ, 1995.
[47] Dror Feitelson and Bill Nitzberg. Job Characteristics of a Production Parallel Scienti c Workload on the NASA Ames iPSC/860. Lecture Nodes on Computer Science, 949, 1995.
[48] Chervenak, A., et al., The Data Grid:Towards an Architecture for the Distributed Management and Analysis of Large Scientific Data Sets. J. Network and Computer Applications, 2001 (23): p.187-200.
[49] N. Darnianou, N.D., E. Lupu, M. Sloman.The Ponder Policy Specification Language. in Policies for Distributed Systems and Networkds. 2001.
[50] Ryutov, T. and C. Neuman, Access Control Framework for Distributed Applications. 1998, IETF. W. Gropp, E. Lusk, N. Doss, and A. Skjellum.A high-performance, portable implementation of the MPI message passing interface standard. Parallel Computing, 22:789-828, 1996..
[51] Korab, H., and Brown, M. D. 1995. Introduction. In Virtual environments and distributed computing at SC'95, New York and Los Alamitos, ACM/IEEE, p. 1-6.
[52] http://www-fp.mcs.anl.gov/dsl/GridFTP-Protocol-RFC-Draft.pdf
[53] R. Arpaci, A. Dusseau, A. Vahdat, L. Liu, T. Anderson, and D. Patterson. The interaction of parallel and sequential workloads on a network of workstations. In Proc. SIGMETRICS, 1995.
[54] Kenneth P. Birman and Robbert van Renesse. Software for reliable networks. Scienti
c American, May 1996.
[55] R. Unrau, O. Krieger, B. Gamsa, and M. Stumm. Hierarchical clustering: A structure for scalable multi-processor operating system design. The Journal of Supercomputing, 9(1/2):105-134, 1995.
[56] S. Zhou, M. Stumm, K. Li, and D. Wortmann. Heterogeneous distributed shared memory (Mermaid). IEEE Transactions on Parallel and Distributed Systems, 3(5):540-554, September 1992.
[57] R. Watson and R. Coyne. The parallel I/O architecture of the high performance storage system (HPSS).In 14th IEEE Symposium Mass Storage Systems, Monterey, CA, September 1995. Comp. Soc. Press.
[58] http://www.21vianet.com/pdf/loadbalance.pdf, 2001
[59] Message Passing Interface Forum. MPI-2: Extensions to the Message-Passing Interface, 1997. http://www.mpi-forum.org
[2] 周培德编著：算法设计与分析．机械工业出版社．2002：308-329页
[3] 黄铠等，可扩展并性计算技术、结构与编程．机械工业出版社．2000：429-458页
[4] 路鑫达等译，并行程序设计．机械工业出版社．2002：31-177页
[5] 都志辉编著．高性能计算并行编程技术．清华大学出版社．2001：38-81页
[6] Rajkumar Buyya著，高性能集群计算：结构与系统(第一卷)．电子工业出版社．2001：354-445
[7] Dale Rogerson著，COM技术内幕．清华大学出版社．2000：27-79页
[8] 邓昕等，多机网络接口的性能提高，96全国计算机体系结构学术研讨会，1996：1-5
[9] 熊建新等，一种分析基于消息通讯模型的并行程序中的不确定性的方法，第六届全国青年工作者会议，1996：69-74
[10] 傅强等，基于PVM的动态任务负载平衡系统的实现，97全国计算机体系结构学术会议，1997：57-62
[11] 杜术，消息传递系统MPI检查点和进程迁移技术的研究与实现，中国学位论文数据库，2001
[12] 刘岩，并行处理与计算机网络中的组合优化问题的研究，中国学位论文数据库，1996
[13] 徐松，分布式系统中的任务迁移和动态负载均衡调整，中国学位论文数据库
[14] 周泽华，消息中间件管理研究，中国学位论文数据库，2000
[15] 李琰，韩焱，对雷达视频图象的压缩处理算法实现．船舶论证参考．2002年第3期．No．3．2002．50-54页
[16] 梁剑等，CORBA技术剖析及其应用．计算机应用研究．2002年第8期．90-93页
[17] 肖龙，实时操作系统在导航显控台中的应用．船舶论证参考．2002年第3期No．3．2002．7-11页
[18] 刘从越，利用UML获取功能需求的方法．计算机应用研究．2002年第8期．88—90页
[19] 周世兵，运用UML为软件项目建模研究．计算机应用研究．2002年第8
期．85-87页
[20]王瑞金等，统一建模语言UML及其建模实例．计算机应用研究．2002年第8期．80-84页
[21]黄汝维等，Delphi中三层结构与CORBA的应用研究计算机应用研究．2002年第8期．94-96页
[22]陈莉群等，linux操作系统内核分析．人民邮电出版社．2000：33-69页
[23]梁玉柱等，COM+技术解决方案设计．机械工业出版社．2001
[24]潘爱民等，Visual C++技术内幕．清华大学出版社．2001
[25]官章全等，Visual C++6.0类库大全．电子工业出版社．1999
[26]吴定一等，异步传递模式的理论与应用．科学出版社
[27]施振川等，UNIX网路编程(第2版)第1卷：套接口API和X./Open传输接口API．清华大学出版社．2001．9
[28]王书洪等，Windows高级编程指南．清华大学出版社．1999
[29]刘宗田等，C++编程思想．机械工业出版社．2001
[30]胡谷雨等，TCP/IP详解卷3．机械工业出版社．2001
[31]陈逸等，DCOM入门．中国电力出版社．2001
[32]潘爱民等，COM本质论．中国电力出版社．2001：33-64页
[33]L.Duchien and L.Seinturier ,OBSERVATION OF DISTRIBUTED COMUTATIONS: A REFLECTIVE APPROACH FOR CORBA,International Journal of Parallel and Distributed Systems and Networks,Vol.4.No. 1,2001 P17-25
[34]Gelff Coulson,Gordon S.Blair,Michael Clarke,Nikos Parlabvantzas ,The design of a configurable and reconfigurable middleware platform, Distributed Computing,(2002) 15:109-126
[35]熊贵喜等，计算机网络(第3版)．清华大学出版社．2000：22-138页
[36]林生，计算机通信网原理．西安电子科技大学出版社．1997
[37]陈向群等，现代操作系统．机械工业出版社．2001
[38]孔祥营等，嵌入式操作系统VxWorks及其开发环境Tornado．中国电力出版社．2002
[39](美)Behrouz Forouzam等著，潘仡等译，数据通信与网络．机械工业出版社．2000
[40]The Globus Project, www.globus.org
[41]K. Czajkowski, I. Foster, C. Kesselman, S. Martin, W. Smith, and S. Tuecke. A resource management architecture for metacomputing systems. Technical report, Mathematics and Computer Science Division, Argonne National Laboratory, Argonne, Ⅲ., 1997. Submitted.
[42] S. Fitzgerald, I. Foster, C. Kesselman, G. von Laszewski, W. Smith, and S. Tuecke. A directory service for conguring high-performance distributed computations. In Proc. 6th IEEE Symp.on High Performance Distributed Computing, pages 365-{375. IEEE Computer Society Press, 1997.
[43] I. Foster, J. Geisler, and S. Tuecke. MPI on the I-WAY: A wide-area, multimethod implementa-tion of the Message Passing Interface. In Proceedings of the 1996 MPI Developers Conference, pages 10-17. IEEE Computer Society Press, 1996.
[44] I. Foster and C. Kesselman, editors. Computational Grids: The Future of High-Performance Distributed Computing. Morgan Kaufmann Publishers, 1998.
[45] W. Gropp and E. Lusk. An abstract device denition to support the implementation of a high-level point-to-point message-passing interface. Preprint MCS-P342-1193, Mathematics and Computer Science Division, Argonne National Laboratory, Argonne, Ⅲ. 1994.
[46] W. Gropp and E, Lusk. MPICH working note: Creating a new MPICH device using the channel interface. Technical Report ANL/MCS-TM-213, Mathematics and Computer Science Division, Argonne National Laboratory, Argonne, Ⅲ, 1995.
[47] Dror Feitelson and Bill Nitzberg. Job Characteristics of a Production Parallel Scienti c Workload on the NASA Ames iPSC/860. Lecture Nodes on Computer Science, 949, 1995.
[48] Chervenak, A., et al., The Data Grid:Towards an Architecture for the Distributed Management and Analysis of Large Scientific Data Sets. J. Network and Computer Applications, 2001 (23): p.187-200.
[49] N. Darnianou, N.D., E. Lupu, M. Sloman.The Ponder Policy Specification Language. in Policies for Distributed Systems and Networkds. 2001.
[50] Ryutov, T. and C. Neuman, Access Control Framework for Distributed Applications. 1998, IETF. W. Gropp, E. Lusk, N. Doss, and A. Skjellum.A high-performance, portable implementation of the MPI message passing interface standard. Parallel Computing, 22:789-828, 1996..
[51] Korab, H., and Brown, M. D. 1995. Introduction. In Virtual environments and distributed computing at SC'95, New York and Los Alamitos, ACM/IEEE, p. 1-6.
[52] http://www-fp.mcs.anl.gov/dsl/GridFTP-Protocol-RFC-Draft.pdf
[53] R. Arpaci, A. Dusseau, A. Vahdat, L. Liu, T. Anderson, and D. Patterson. The interaction of parallel and sequential workloads on a network of workstations. In Proc. SIGMETRICS, 1995.
[54] Kenneth P. Birman and Robbert van Renesse. Software for reliable networks. Scienti
c American, May 1996.
[55] R. Unrau, O. Krieger, B. Gamsa, and M. Stumm. Hierarchical clustering: A structure for scalable multi-processor operating system design. The Journal of Supercomputing, 9(1/2):105-134, 1995.
[56] S. Zhou, M. Stumm, K. Li, and D. Wortmann. Heterogeneous distributed shared memory (Mermaid). IEEE Transactions on Parallel and Distributed Systems, 3(5):540-554, September 1992.
[57] R. Watson and R. Coyne. The parallel I/O architecture of the high performance storage system (HPSS).In 14th IEEE Symposium Mass Storage Systems, Monterey, CA, September 1995. Comp. Soc. Press.
[58] http://www.21vianet.com/pdf/loadbalance.pdf, 2001
[59] Message Passing Interface Forum. MPI-2: Extensions to the Message-Passing Interface, 1997. http://www.mpi-forum.org