基于petri网的多处理机通信协议的性能分析及评估-计算机软件及理论专业论文.docxVIP

擅要P 擅要 P 461 6 7．耋 摘要 I随着计算机硬件的发展，网络硬件支持每秒钟大约为Gb／s的带宽已成 为习读。但是用户应用程序所使用的带宽却一直相对低些，并且应用程序之 间的通信延迟也相对高些。其中最主要的原因恐怕就是分层网络软件频繁的 系统调用和内存．内存拷贝，造成了物理层和应用层之间的性能鸿沟。虚拟 接13体系结构是一种崭新的机群系统内部通信思想，它为用户提供了一个直 接与网络相联系的虚拟接口，抛弃了传统通信协议的多层协议体药使操作 系统不参与进程双方的通信过程，这样就大大提高了通信效率习本文首先介 绍了Petri网的研究概况，指出利用Petri网进行通信协议的描述和验证的优 越性。随后，在Petd网基本概念和相关性质的基础上，对停．等协议建立了 Petri网模型，得到相关结论。在引入虚拟接口体系结构的基础上，为Berkeley 模型建立了Pelri网模型，通过模型对其工作模式进行详细描述，并对其中 存在的活锁问题给出了具有实际意义的解决方法。最后，借助模型的可达图 得到马尔可夫链，由马尔可夫过程建立方程组得到相关的性能参数指标，为 今后的研究创造了理论条件。 关键字：Petri网，马尔可夫链，马尔可夫过程，用户层网络，允梅，通 信协议，虚拟接口体系结构 AbstractWith Abstract With the development of computer hardware，it has been rea／ized that the balldwidth has reached gigabit per second supporting by the network hardware．At the same time．the baIld丽dm of applications is somehow lower．The main reason is the continually system call and memory-to·memory copy,causing the performance chasm between the physical level and application level．The virtual interface architecture is a new communicating idea among the cluster system．It provides a virtual interface between applications and the ne“vork to US． abandoning multiple-level protocol system of the traditional communicating protocols+In this system，the operating system doesn’t participate in the communicating procedure between the bom processes．thus advancing the communicating rate．In this paper,we first introduced the research survey in Petri nets．T11is shows the advantage of using Petri nets to describe and veilfy communicating protocols．Soon afterwards we construct the model of stop-and— wait protocol witll the tool of automatic machine and Petri nets．After introducing the virtual interface architecture．we construct the model of Berkeley model诚tll Petri nets，through which we put the solvent oflive lock in the model．At last，we have the Markov chain by use of the reachable picture．On the base of Markov chain，we construct the linear equations through Markov procedure．So we provide the performance targets to the future research work． Key