一种基于随机进程代数的软件性能预测方法  

Software Performance Prediction Approach Based on Stochastic Process Algebra

作　　者：任正平[1] 刘晓明[1] 黄松[1] 叶旭光[2] 

机构地区：[1]解放军理工大学指挥自动化学院军用软件测评中心,江苏南京210007 [2]解放军理工大学指挥自动化学院作战软件教研室,江苏南京210007

出　　处：《系统仿真学报》2008年第15期3996-3999,4024,共5页Journal of System Simulation

摘　　要：性能是软件系统的内在属性之一,主要取决于体系结构选择和软件设计。基于模型的性能预测方法可以在软件开发完毕前得到性能需求满足情况的评价,也可帮助系统进行性能规划和调优。使用随机进程代数PEPA对J2EE应用交互过程进行了建模,该模型描绘了远程调用过程和容器处理过程的细节,能有效反映分布式应用中影响性能的主要因素。基于PEPA模型提出了一种面向COTS系统的性能预测方法,并可用于指导改进系统的响应性和可伸缩性。通过一个J2EE实例验证了方法的有效性。Performance is one of the inherent attributes of software system. It is mainly decided by the selection of software architecture and designing information. Whether the performance demands are meet can be evaluated by model-based performance prediction technologies, So do the capability projecting and the fine-tuning. The interaction process of J2EE was modeled using Performance Evaluation Process Algebra （PEPA）. Remote invocation and container processing details are represented in the model, and major performance factors of distributed J2EE application are reflected. A COTS-oriented performance prediction approach based on PEPA models was proposed. It can help to improve responsibility and scalability besides predicting system performance. The approach was verified by a J2EE application case study.

关 键 词：性能预测 性能模型 COTS PEPA J2EE应用 

分 类 号：TP302.7[自动化与计算机技术—计算机系统结构] TP311.5[自动化与计算机技术—计算机科学与技术]