Software Performance Evaluation Based On UML And CPN Research

Because of the impact of traditional conception, software performance is always managed after the progress of software development finished. If defacts are found, we con't choose but modify from the beginning of the project. This method maybe results in all the software design overthrowed, which is not applied for the large-scale system. Software performance evaluates software qualitity from the beginning of software design, so it can find and modify software defact as early as possible and improve the effectiveness of software production.Software performance evaluation has gone development process from the early method of measurement methods to model method. Because analytical model often limits to the input of system and the method of mathematical theory is lack of flexibility, it is a new direction to create model symbols. UML becomes the untified modeling language on the process of software implementation. A lot of performance evaluation methods start from UML model. UML is not a formal language and lack of formal semantics, so it can't simulate the system dynamically. Colored Petri net is a kind of high-level Petri net, and it can model the complex system with hierarchical models and identify the source of the system with colset. Also it can simulate the system model dynamically. So CPN is the foothold of software performance starting with UML.Software performance based on UML and CPN bases on UML2.0 entirly. First of all it models the system using UML2.0. And capture system requirement using use case diagram and refine the use case using activity diagram. Then the transformation rules from UML2.0 model to CPN model is proposed. The use case diagram is transformed into CPN top model and activity diagram is transformed to the corresponding sub-networks, then all the CPN models is connected to the entire system. Lastly the CPN model performance is evaluated by the mechanism of monitor and state space, such as system throughput, waiting time and processing time of tasks.This approach does not not only limit the distribution of system input but also expand the UML model. And it can simulate the system model dynamically. The software performance can be detected during the simulation. Also the intuitive model is beneficial to understand and confirme requirement of user. The results of performance evaluation can be used to find the most optimal UML model design, which can fulfill the requirement of performance in the forepart of software development.