Early performance prediction of SDL/MSC specified systems by automated synthetic code generation

We present a new approach for early performance prediction based on MSC specified systems in the context of SDL. Our approach is integrated into existing design methodologies as proposed by commercial tool vendors where communication software is fully specified in SDL and the final implementation is derived from there. Obviously the structure of the SDL specification will influence the performance of the final system. Thus it is very important to make performance estimates for the target system available at design time to steer important design decisions. The key to performance evaluation is the development of performance models, which are evaluated either with analytical, simulation or monitoring techniques. Our approach is scenario-based and uses non-functional annotations of MSCs to formalise additional performance attributes. These MSCs are automatically transformed to an SDL specification which yields a prototype implementation via a code generation tool chain. The resulting implementation is executed on the target machines with the target system software from which many performance characteristics can be evaluated using monitoring techniques. We call the implementation “synthetic” since it is artificially derived from MSC specifications and the generated SDL is not designed to be subsequently reused through functional refinement. This paper describes the automatic transformation from MSC to the SDL specification and its environment with a special focus on dynamic issues.