| With the software scale and complexity increasing, software products are faced with a large number of concurrent users and computing operations. So the software performance is more and more concerned. In this case, stress testing, which is implemented to verify system performance, has become particularly important. Stress testing as a method of software testing can test software used in peak performance by simulating a huge workload. So it can find the performance shortcomings of software as soon as possible to prevent the losses caused by blindly putting the products on the market.With complexity and uncertainty of stress testing, the accuracy of stress testing can be affected by many factors. Therefore, we need to improve the accuracy of stress testing for software performance analysis. In this paper, we have done an in-depth study and lots of practical work about the testing process model, testing tool design, testing parameter estimation methods, concurrency control, performance monitoring and other aspects of stress testing. With the knowledge of automated software testing, stochastic processes and queuing theory, we have proposed some methods to improve the accuracy of stress testing, and proposed a software reliability model based on the analysis of results from stress testing. The study in this paper has a certain depth and innovation. The research results can be used to practical stress testing work well.This paper has also designed and implemented a concurrency tool named MCT (Mini Concurrency Tool) for stress testing. MCT can support the concurrent testing on HTTP protocol and control the collection point on many load generators.It uses XML which is easy to prepare as test script. It has large amount of simulation concurrency, high implementation speed, less occupying of system resources. Finally, this paper presented an example of stress testing project, with the theory studied in this paper and the testing tool MCT designed in this paper. |
PDF Full Text Request