Agent-Aided Software Engineering of High Performance Applications

High performance applications are parallel and distributed applications whose performance is usually the most concerned by the developers. Traditional software tools of performance evaluation can not be used efficiently in the Computational Grid, a new software and hardware infrastructure that provides inexpensive access to highend computational capabilities [3], because software developers of grid-enabled high performance applications may have not much knowledge on parallel and distributed computing and may be not so professional to use current performance modelling and evaluation tools. In this work the concept of Agent-Aided Software Engineering (AASE) is presented to overcome these difficulties, which integrates autonomous agents with the CASE tools to aid application developers to cope with not only the complexity but also the professionality in software development processes. The Performance Evaluation Agent (PEA) can act autonomously as an expert in the grid application development environment to assist and guide the software development. The main parts of the PEA include an Abstractor, an Evaluator, and an Advisor.