Automatic Generation Of Workload For Web2.0Application

Automatic generation of workload is very important for Web2.0applications.Compared with Web1.0application, Web2.0application is a new application form, whichhas a larger number of users, more kinds of user types, richer user experiences, morecomplex user actions and user interactions, more kinds of user scenarios than Web1.0applications.However, with more complex system under test (SUT), old design ideas are stillused in the design process of test plans. Artificial heuristic methods are still used in thedesign process of workload. When testers test the performance of an Internet application(such as a website), the method they preferred is to use a load generation tool to generateHTTP requests from a low concurrency to a high concurrency to measure theconcurrency capacity of this application. This approach is simple to design and effectivein the circumstance of Web1.0application, so it is widely used by testers. However, withthe rapid development of the Web2.0applications, testing requirements of applicationsbecome more complex. In this circumstance, testers have to confirm each feature orcomponent of SUT, design testing model, set the parameters of the model, design testload, and get final test results. Obviously, the testing process is very complex. Moreover,because of the lack of supportive data when setting the parameters for the model, testresults are lack of credibility. Therefore, this method is no longer applicable and efficient.Also, in the face of numerous types of Web2.0applications, there is no an effective,flexible test plan to generate automatically a workload for Web2.0applications.To solve this problem, the paper designed an effective solution to automaticallygenerated workload for Web2.0applications. The solution focused on the automaticgeneration of SNS (social network service) Web2.0applications workload, which canalso be extended to other types of Web2.0applications as well. The works of the solutionin this paper include: analyzing of Web2.0application characteristics, collecting user data from Web2.0application, analyzing and modeling user data of Web2.0application,designing and automatically generating Web2.0application workload.First, the paper analyzed Web2.0application characteristics and designed theworkload of Web2.0application by two levels. The two level loads are single-userbehavior load and concurrent user load. The design and implementation process of userconcurrent load includes: collecting Web2.0application user concurrency data byautomated means, designing four kinds of concurrent loads according to the analysis ofuser concurrency, implementing a module that can generate concurrent load dataautomatically on Faban. The design and implementation process of single-user behaviorload includes: analyzing user data extracted from a Web2.0application background log,modeling user behavior based on Web2.0application user data, using an algorithm basedon Markov chain to analyze Web2.0user behavior model, designing a load according tothe analyzing results. Then, the paper combined the two loads into a uniform workloadthat can both meet the needs of dynamic changes in concurrent loads and thecharacteristics of real user behaviors. Finally, in order to demonstrate the solution, thepaper did an experiment to implement the workload, collect test data and analyze testresults.