Research Of Load Model For Cloud-based Web Performance Testing

With the development of Internet,web application system is widely used in society.System performance is becoming more and more important.Performance testing is used to determine metrics(such as response time,throughput,and resource utilization)of enterprise system for evaluating.Convinced test results need to be guaranteed by realistic load scenarios.Realness of load scenarios mainly depends on two factors: test environment and test method.Starting from these two factors,this thesis focuses on how to provide realistic load scenarios for target system in performance testing,and then to ensure that the collected performance indicators can reflect characteristics of the system in production.This thesis presents a conceptual performance testing framework by applying Cloud Compute to performance testing.This framework contains several modules: computing,storage,message queues,and test platform.As the core module of this framework,testing platform contains a variety of functional engines,which can enhance the level of testing automation.This platform can be deployed in the enterprise data center or the cloud.This framework can flexibly use cloud resources to reduce business costs,and can provide a more realistic and reliable test environment with massive resources and multi-regional data centers.Then this thesis focuses on test methods,and a novel load model suitable for cloud environment is designed.This load model is composed of two parts: user behavior modeling method and C/S architecture load tool.The user behavior modeling method can be used to generate scenarios for load tool,which can automatically establish user behavior model by analyzing Web logs.Based on user behavior model,the representative scenario cases can be generated.The load tool is composed of server side and client side.Clients receive test task assigned by the server,then drive real browsers to send load requests to enterprise system.Finally,the cloud load model(including C/S architecture load tool and user behavior modeling method)is implemented,and then it is applied to a case analysis.The experimental results show that the load model meets flexible requirement for using in cloud,which also ensures the realness of load scenarios that system undertaken.The load tool uses real browser as load client,so it can effectively solve the shortage of the traditional testing tools,such as not execute JavaScript and not simulate Ajax asynchronous interaction.The user behavior modeling method can realistically reproduce user behavior of production scenario,can reduce the randomness of selecting scenario cases,and thus can provide representative scenario cases for performance testing.