集群服务器系统建模研究及性能评价
摘要
随着网络的普及,网络服务器需要为越来越多的用户提供服务,在这种情况下,即使单台服务器性能再高,所能提供的服务也是有限的,而且其工作量很容易在访问高峰时期过载。集群服务器结构能较好地解决上述问题,因此,研究集群服务器系统以及其负载均衡策略具有一定的理论意义和实用价值。
论文从两个方面研究了集群服务器系统的性能,一是它的可靠性或可利用性;另一方面是它的吞吐率、响应时间、利用率等处理效率。阐述了Petri网的主要特性:并行、不确定性、异步和分布描述能力以及分析能力等。展示了Petri网除具有类似流程图、框图和网图的可视描述功能外,还可通过标记的流动模拟系统的动态、活动行为。论文以广东某公司的“分布式系统性能监测及反馈控制技术”项目为背景,在原项目单纯地从监测服务器负载大小来决定终端是否继续进行新请求的思想基础上,提出了在客户端与服务器端增添请求分配器的方法,将后端互为备用的服务器架构成集群结构。在此集群服务器系统中,将后端服务器所采用的请求调度优先级策略与请求分配器所使用的复杂均衡策略相结合的综合控制思想,实现负载均衡和QoS控制。并且采用随机高级Petri网性能功能模型与分析技术,对QoS-aware负载均衡策略的性能优越性进行了分析和研究。
论文研究对象是负载均衡策略在集群服务器系统中的优化应用。在对系统的建模,模型仿真和性能评价方面所做的研究工作,对于同类系统建模及性能评价具有较好的借鉴作用。
With the popularization of the network, the network servers have to provide more and more services for clients. Even the performance of a single server is high, the service which is provided by it is very limited. In visit peak period, the server is easily overloaded. Therefore, the cluster server system produced.
As the ordinate computer system, the performance of the cluster server system includes two aspects: one is its reliability and usability, the other is its efficiency. There are three types included in efficiency. They are throughput, response time and utilization. Petri net is a useful graphics and math modeling tool which can be used in many systems and fields. Its characteristics include concurrency, nondeterminacy, asynchronism and distributed ability to describe and analyze. Petri nets have the functions as flowchart diagram and network map which can describe with visible method. It also can describe dynamic abilities of simulation systems with‘token’.
The thesis’s background is the project named“the Performance Inspection and Feedback Control of Web-Service System”of a company. The thesis proposes integrated control idea, which adopts complex balancing strategy used in request distributor cooperating with priority of request used in servers. The thesis models cluster servers system based on the theory and technique of stochastic high-level Petri net, then analysis the performance of system.
The research is to analysis and simulate the performance of the cluster server system, which is modeled with Petri net. This implements load balancing and QoS control of the system. It is very helpful for the related researches.
论文从两个方面研究了集群服务器系统的性能,一是它的可靠性或可利用性;另一方面是它的吞吐率、响应时间、利用率等处理效率。阐述了Petri网的主要特性:并行、不确定性、异步和分布描述能力以及分析能力等。展示了Petri网除具有类似流程图、框图和网图的可视描述功能外,还可通过标记的流动模拟系统的动态、活动行为。论文以广东某公司的“分布式系统性能监测及反馈控制技术”项目为背景,在原项目单纯地从监测服务器负载大小来决定终端是否继续进行新请求的思想基础上,提出了在客户端与服务器端增添请求分配器的方法,将后端互为备用的服务器架构成集群结构。在此集群服务器系统中,将后端服务器所采用的请求调度优先级策略与请求分配器所使用的复杂均衡策略相结合的综合控制思想,实现负载均衡和QoS控制。并且采用随机高级Petri网性能功能模型与分析技术,对QoS-aware负载均衡策略的性能优越性进行了分析和研究。
论文研究对象是负载均衡策略在集群服务器系统中的优化应用。在对系统的建模,模型仿真和性能评价方面所做的研究工作,对于同类系统建模及性能评价具有较好的借鉴作用。
With the popularization of the network, the network servers have to provide more and more services for clients. Even the performance of a single server is high, the service which is provided by it is very limited. In visit peak period, the server is easily overloaded. Therefore, the cluster server system produced.
As the ordinate computer system, the performance of the cluster server system includes two aspects: one is its reliability and usability, the other is its efficiency. There are three types included in efficiency. They are throughput, response time and utilization. Petri net is a useful graphics and math modeling tool which can be used in many systems and fields. Its characteristics include concurrency, nondeterminacy, asynchronism and distributed ability to describe and analyze. Petri nets have the functions as flowchart diagram and network map which can describe with visible method. It also can describe dynamic abilities of simulation systems with‘token’.
The thesis’s background is the project named“the Performance Inspection and Feedback Control of Web-Service System”of a company. The thesis proposes integrated control idea, which adopts complex balancing strategy used in request distributor cooperating with priority of request used in servers. The thesis models cluster servers system based on the theory and technique of stochastic high-level Petri net, then analysis the performance of system.
The research is to analysis and simulate the performance of the cluster server system, which is modeled with Petri net. This implements load balancing and QoS control of the system. It is very helpful for the related researches.
引文
[1] Commoner F, Holt A W, Even S, et al. Marked directed graphs. J. Comput. Syst. Sci., 1971, 5: 511~523
[2] Peterson J L. Petri Net Theory and the Modeling of systems. Englewood Cliffs, NJ: Prentice-Hall, Inc., 1981
[3]林闯.计算机网络和计算机系统的性能评价.北京:清华大学出版社, 2001. 155~169
[4]丁原,刘玉树,朱天缓.集群服务器技术综述.兵工自动化, 2000, 3: 10~13
[5]张志友.计算机集群技术概述,实验室研究与探索. 2006, 25(5): 607~609
[6]吴立德,吴霭成.计算机系统性能评价.上海:上海科学技术出版社, 1986. 7~8
[7]袁崇义. Petri网原理.北京:电子工业出版社, 2005. 3~95
[8]付国为,刘心松,冯玮.基于副反馈的网络负载调度算法.成都信息工程学院学报, 2006, 21(6): 779~782
[9]黄敏,姚正林,刘金刚.网络QoS调度策略的分析与研究.计算机工程与应用, 2006, 42(29): 118~120
[10]刘好胜,张颖江,贺红艳.构建安全的负载均衡集群系统.科技资讯, 2006, 11: 21~22
[11]周丽,董蔚,宋义.集群系统下的动态网络负载平衡.电测与仪表, 2005, 42(1): 26~28
[12]罗秉安,张立臣.集群系统的技术与应用,微电脑应用. 2003, 19(1): 62~64
[13]梁晓珺.基于着色Petri网的QoS驱动服务合成建模及应用.中山大学研究生学刊(自然科学、医学版), 2006, 26(1): 121~132
[14]集群下的Web QoS控制. http://www.ccw.com.cn/htm/center/tech/02_11_1_2.asp. 2002-11-1
[15] van der Aalst WMP. The application of Petri nets to workflow management. Journal of Circuits, Systems and Computers, 1998, 8(1): 21~66
[16]袁志祥.基于时间Petri网的Web服务建模技术研究.计算机工程与设计, 2006, 27(20): 3801~3804
[17]林闯,李雅娟,单志广.基于随即Petri网的系统性能评价.清华大学学报(自然科学版), 2003, 43(4): 475~479
[18] Henderson W, Taylor P G. Embedded processes in stochastic Petri nets. IEEE Trans on Software Eng, 1991, 17(2):108~116
[19] Jungnitz H, Sanchez B, Silve M. Approximate throughput computation of stochastic marked graphs . IEEE Tran Software Engineering, 1994, 20(7): 526~535
[20] Buchholz P, Ciardo G, Donatelli S, et al. Complexity of memoty-efficient Kronecker operations with applications to the solution of Markov models. INFOMS Journal on Computing, 2000, 12(3): 203~222
[21] Deavours D, Clark G, Courtney T, et al. The Mobius framework and its implementation. IEEE Trans on Software Engineering, 2002, 28(10): 956~969
[22] Boucherie R J. A characterization of independence for competing Markov chains with applications to Stochastic Petri nets. IEEE Trans Software Engineering, 1994, 20(7): 536~544
[23] ShaikhAli A, Rana O, Al-Ali R, et al. UDDIe: An Extended Registry for Web Services. Proc. of the Service Oriented Computing: Models, Architectures and Applications, 2003, 2(1): 85-89
[24] Wang Y, Hu CM, Huai JP. A new grid workflow description language. In: Proc. of the 2005 IEEE Int’l Conf. on Services Computing, Vol2. Orlando: IEEE Computer Society, 2005, 2(14): 257~258
[25] Ran SP. A model for Web services discovery with QoS. ACM SIGecom Exchanges, 2003, 4(1): 1~10
[26] Grefen P, Aberer K, Hoffner Y, Ludwig H. CrossFlow: Cross-organizational workflow management in dynamic virtual enterprises. Int’l Journal of Computer Systems Science & Engineering, 2000, 15(5):227~290
[27] Buyya R, Abramson D, Venugopal S. The grid economy. Proc. of the IEEE, 2005, 93(3): 698~714
[28] T W Kuo, W R Yang, K J Lin. A Class of Rate-Based Real-Time Scheduling Algorithms.IEEE Trans on Computers, 2002, 51(6): 708~720
[29] G Xue. Minimum-Cost QoS Multicast and Unicast Routing in Comunication Networks. IEEE Trans on Communications, 2003, 51(5): 817~824
[30] M Charka, J Nao, B Schieber. Resource Optimization in QoS Multicast Routing of Real-Time Multimedia. IEEE/ACM Trans on Networking, 2004, 12(2): 340~348
[31]伍之昂,罗军舟,宋爱波.基于QoS的网格资源管理.软件学报, 2006, 17(11): 2264~2276
[32]金海,陈汉华. CGSP作业管理器合成服务的QoS优化模型及求解.计算机学报, 2005, 28(4): 578~588
[33]朱穗晖,都志辉,林伟坚.网格环境合同计算元服务的设计与应用.计算机学报, 2005, 28(4): 486~494
[34]岳昆,王晓玲,周傲英. Web服务核心支撑技术.软件学报, 2004, 15(3): 428-442
[35]崔建群,吴黎兵,彭熙,等.支持QoS属性的Web服务PFS模型研究.计算机工程, 2006, 32(21), 100~102
[36]王建锋,钱雪忠.基于中间件的Web服务和网格服务混合合成.微计算机信息, 2006, 22(7), 86~88
[37]王重钢,隆克平.分组交换网络中队列调度算法的研究及其展望.电子学报, 2001, 29(4): 553~559
[38]王宏宇,顾冠群.集成服务网络中的分组调度算法综述.计算机学报, 1999, 22(10): 1090~1099
[39]张普,王青,杨立光.网络计算机集群负载均衡机制的研究.计算机工程与设计, 2006, 27(16): 2914~2917
[40]郭国强. IP流量规划对网络延迟性能的影响.常德师范学院学报, 2001, 13(2): 91
[41]吴辰文.基于响应时间的LAN系统性能评价.兰州交通大学学报(自然科学版), 2004, 23(1): 68~70
[42]夏明波,王晓川,金士尧.自调度集群的研究与实现.重庆邮电学院学报(自然科学版), 2006, 18(6): 748~751
[43]周双娥,刘艳.基于Petri网的信息系统建模及性能分析.微型电脑应用, 2006, 22(6):26~28
[44]童蕾,王宏安,戴国忠.基于Petri网的Web服务流程建模方法研究.计算机仿真, 2006, 23(1): 69~73
[45]张文杰,钱德沛,张然,等.互联网应用性能测量系统的研究与实现.计算机研究与发展, 2003, 40(1): 60~67
[46]周胜源,敖发良,陈名松. RPR三种优先级业务的时延吞吐量性能分析.电讯技术, 2005, 4: 28~33
[47] Lin Chuang, Yang Shiqiang. Performance analysis of scheduling schemes in multiserver multiqueue systems. Acta Electronic Sinica, 2000, 28(5): 17~20
[48]曲绍刚,杨广文,林闯,等.基于完成时间的任务分配方案与性能分析.计算机研究与发展, 2005, 42(8): 1397~1402
[2] Peterson J L. Petri Net Theory and the Modeling of systems. Englewood Cliffs, NJ: Prentice-Hall, Inc., 1981
[3]林闯.计算机网络和计算机系统的性能评价.北京:清华大学出版社, 2001. 155~169
[4]丁原,刘玉树,朱天缓.集群服务器技术综述.兵工自动化, 2000, 3: 10~13
[5]张志友.计算机集群技术概述,实验室研究与探索. 2006, 25(5): 607~609
[6]吴立德,吴霭成.计算机系统性能评价.上海:上海科学技术出版社, 1986. 7~8
[7]袁崇义. Petri网原理.北京:电子工业出版社, 2005. 3~95
[8]付国为,刘心松,冯玮.基于副反馈的网络负载调度算法.成都信息工程学院学报, 2006, 21(6): 779~782
[9]黄敏,姚正林,刘金刚.网络QoS调度策略的分析与研究.计算机工程与应用, 2006, 42(29): 118~120
[10]刘好胜,张颖江,贺红艳.构建安全的负载均衡集群系统.科技资讯, 2006, 11: 21~22
[11]周丽,董蔚,宋义.集群系统下的动态网络负载平衡.电测与仪表, 2005, 42(1): 26~28
[12]罗秉安,张立臣.集群系统的技术与应用,微电脑应用. 2003, 19(1): 62~64
[13]梁晓珺.基于着色Petri网的QoS驱动服务合成建模及应用.中山大学研究生学刊(自然科学、医学版), 2006, 26(1): 121~132
[14]集群下的Web QoS控制. http://www.ccw.com.cn/htm/center/tech/02_11_1_2.asp. 2002-11-1
[15] van der Aalst WMP. The application of Petri nets to workflow management. Journal of Circuits, Systems and Computers, 1998, 8(1): 21~66
[16]袁志祥.基于时间Petri网的Web服务建模技术研究.计算机工程与设计, 2006, 27(20): 3801~3804
[17]林闯,李雅娟,单志广.基于随即Petri网的系统性能评价.清华大学学报(自然科学版), 2003, 43(4): 475~479
[18] Henderson W, Taylor P G. Embedded processes in stochastic Petri nets. IEEE Trans on Software Eng, 1991, 17(2):108~116
[19] Jungnitz H, Sanchez B, Silve M. Approximate throughput computation of stochastic marked graphs . IEEE Tran Software Engineering, 1994, 20(7): 526~535
[20] Buchholz P, Ciardo G, Donatelli S, et al. Complexity of memoty-efficient Kronecker operations with applications to the solution of Markov models. INFOMS Journal on Computing, 2000, 12(3): 203~222
[21] Deavours D, Clark G, Courtney T, et al. The Mobius framework and its implementation. IEEE Trans on Software Engineering, 2002, 28(10): 956~969
[22] Boucherie R J. A characterization of independence for competing Markov chains with applications to Stochastic Petri nets. IEEE Trans Software Engineering, 1994, 20(7): 536~544
[23] ShaikhAli A, Rana O, Al-Ali R, et al. UDDIe: An Extended Registry for Web Services. Proc. of the Service Oriented Computing: Models, Architectures and Applications, 2003, 2(1): 85-89
[24] Wang Y, Hu CM, Huai JP. A new grid workflow description language. In: Proc. of the 2005 IEEE Int’l Conf. on Services Computing, Vol2. Orlando: IEEE Computer Society, 2005, 2(14): 257~258
[25] Ran SP. A model for Web services discovery with QoS. ACM SIGecom Exchanges, 2003, 4(1): 1~10
[26] Grefen P, Aberer K, Hoffner Y, Ludwig H. CrossFlow: Cross-organizational workflow management in dynamic virtual enterprises. Int’l Journal of Computer Systems Science & Engineering, 2000, 15(5):227~290
[27] Buyya R, Abramson D, Venugopal S. The grid economy. Proc. of the IEEE, 2005, 93(3): 698~714
[28] T W Kuo, W R Yang, K J Lin. A Class of Rate-Based Real-Time Scheduling Algorithms.IEEE Trans on Computers, 2002, 51(6): 708~720
[29] G Xue. Minimum-Cost QoS Multicast and Unicast Routing in Comunication Networks. IEEE Trans on Communications, 2003, 51(5): 817~824
[30] M Charka, J Nao, B Schieber. Resource Optimization in QoS Multicast Routing of Real-Time Multimedia. IEEE/ACM Trans on Networking, 2004, 12(2): 340~348
[31]伍之昂,罗军舟,宋爱波.基于QoS的网格资源管理.软件学报, 2006, 17(11): 2264~2276
[32]金海,陈汉华. CGSP作业管理器合成服务的QoS优化模型及求解.计算机学报, 2005, 28(4): 578~588
[33]朱穗晖,都志辉,林伟坚.网格环境合同计算元服务的设计与应用.计算机学报, 2005, 28(4): 486~494
[34]岳昆,王晓玲,周傲英. Web服务核心支撑技术.软件学报, 2004, 15(3): 428-442
[35]崔建群,吴黎兵,彭熙,等.支持QoS属性的Web服务PFS模型研究.计算机工程, 2006, 32(21), 100~102
[36]王建锋,钱雪忠.基于中间件的Web服务和网格服务混合合成.微计算机信息, 2006, 22(7), 86~88
[37]王重钢,隆克平.分组交换网络中队列调度算法的研究及其展望.电子学报, 2001, 29(4): 553~559
[38]王宏宇,顾冠群.集成服务网络中的分组调度算法综述.计算机学报, 1999, 22(10): 1090~1099
[39]张普,王青,杨立光.网络计算机集群负载均衡机制的研究.计算机工程与设计, 2006, 27(16): 2914~2917
[40]郭国强. IP流量规划对网络延迟性能的影响.常德师范学院学报, 2001, 13(2): 91
[41]吴辰文.基于响应时间的LAN系统性能评价.兰州交通大学学报(自然科学版), 2004, 23(1): 68~70
[42]夏明波,王晓川,金士尧.自调度集群的研究与实现.重庆邮电学院学报(自然科学版), 2006, 18(6): 748~751
[43]周双娥,刘艳.基于Petri网的信息系统建模及性能分析.微型电脑应用, 2006, 22(6):26~28
[44]童蕾,王宏安,戴国忠.基于Petri网的Web服务流程建模方法研究.计算机仿真, 2006, 23(1): 69~73
[45]张文杰,钱德沛,张然,等.互联网应用性能测量系统的研究与实现.计算机研究与发展, 2003, 40(1): 60~67
[46]周胜源,敖发良,陈名松. RPR三种优先级业务的时延吞吐量性能分析.电讯技术, 2005, 4: 28~33
[47] Lin Chuang, Yang Shiqiang. Performance analysis of scheduling schemes in multiserver multiqueue systems. Acta Electronic Sinica, 2000, 28(5): 17~20
[48]曲绍刚,杨广文,林闯,等.基于完成时间的任务分配方案与性能分析.计算机研究与发展, 2005, 42(8): 1397~1402