Design of BDI Agent for Adaptive Performance Testing of Web Services

Abstract

As services are dynamic discovered and bound in the open Internet environment, testing has to be exercised continuously and online to verify and validate the continuous changes and to ensure the quality of the integrated service-based system. During this process, testing strategies have to be adapted in accordance to the changes in the environment and target systems. Software agents are characterized by context awareness, autonomous decision making and social collaboration capabilities. The paper introduces the design of BDI (Believe-Decision-Intention) agents to facilitate adaptive performance testing of Web Services. The BDI model specifies the necessary test knowledge, test goal and action plan to carry out test and adaptive schedule. Performance testing is defined as a scheduling problem to select the workload and test cases in order to achieve the goal of performance abnormal detection. A two-level control architecture is built. At the TR (Test Runner) level, the BDI agents control the workload of concurrent requests. At the TC (Test Coordinator) level, the BDI agents control the complexity of test cases. Agents communicate and collaborate with each other to share knowledge and test plan. The paper introduces the design of the BDI model, the adaptation rules and the control architecture. Case study is exercised to illustrate the adaptive testing process based on the design of BDI agents.