资源敏感的分布式系统性能建模方法

Performance Modeling Approach for Resource Sensitive Distributed Systems

早期的性能预测可帮助设计人员有效地评估和改进系统设计,降低性能风险,提高软件制品满足性能需求的可信程度。但复杂低效的性能模型构造过程,严重阻碍了它在软件开发过程中的应用。为简化建模复杂度,以典型的UML模型为基础,研究了一种面向分布式系统的性能模型方法,并提出了一种中间模型——资源场景模型(Resource Scenario Model,RSM),解决了多UML视图转化为多性能模型的问题,使设计人员可以根据自身偏好选择熟悉的工具进行性能建模和性能分析。最后,以UML活动图和序列图到分层排队网和通用随机Petri网模型的转换为例,说明了本方法的可行性和有效性。

Performance prediction of software at early stage makes benefit to improve the quality of designs and reduce the performance problems. However, the huge costs of performance modeling approaches make it hnpossible to integrate the approach into the software development processes. In order to develop a cost effective modeling approach, we used the widely accepted UML model to support our work for distributed systems, and designed an intermediate model named Resource Scenario Model（RSM）for multi architecture model to multi performance model transformations. There-fore, designers can use their preferred tools to build and analyze software＇ s performance. To illustrate the effectiveness of our method, we given a case study, in which UML models can be easily transformed into performance models, such as LQN and LGSPN.