从UML到GSPN的转换和性能分析方法

Transforming UML to GSPN for Performance Analysis

UML模型一般不能直接进行性能分析,需要利用模型转换的方法将其转换成其他分析模型,比如排队论、随机进程代数或者随机Petri网等模型。利用Eclipse平台上的Papyrus建立3种类型的UML模型(用例图、部署图和活动图)来对系统进行建模,并利用MARTE规范添加一些性能相关的信息;然后利用ATL实现UML模型到广义随机Petri网(GSPN)模型的转换,并使用XStream将上一步得到的GSPN模型转换成分析工具所支持的格式;最后利用基于GSPN的性能分析方法进行系统性能分析。同时给出了一系列性能指标的计算方法,如利用率、吞吐量、平均等待请求的数目以及响应时间等,可以考察系统性能的多个方面,方便系统设计和开发人员对系统性能进行分析和优化。

An UML model cannot be analyzed for performance requirements directly,and it should be transformed into analyzable models such as queueing models,stochastic process algebra models or stochastic Petri nets models.In this paper,three kinds of UML models（use case diagrams,deployment diagrams and activity diagrams）and suitable annotations from the profile for MARTE were chosen to build performance models by the tool Papyrus on the platform Eclipse.UML models are transformed into GSPN models by ATL,and the obtained GSPN models are further transformed into the formats that analyzers can support.At last,the performance can be analyzed by using the performance analysis method based on GSPN.Some performance metrics are given to investigate the system,including utilization,throughput,the average number of waiting requests and response time,which can be referred by system designers and develo-pers to analyze and optimize the performance.