1 A Novel Approach for Performance Improvement of Multi-Agent based System Architecture

In software engineering, it is important issue that architects estimate the performance of software in the early phase of development process. Many approaches translate architecture based on UML into analytical models. However, in agent based system development, these approaches ignore or simplify the crucial details of the underlying the performance of an agent platform. Additionally, architects have only limited kinds of solution to improve performance of system architecture. In this paper, we propose analytical approach for performance improvement of multi-agent based system. This approach avoids the need for a prototype implementation since architects can determine overall form of the performance equation from the architectural design description and can enhance the system architecture to derive optimal architecture from analytical model. We verify the effectiveness of our approach through U-Commerce scenario. 1