Automatically building service evaluation metadata in a grid environment

This paper presents a mechanism that automatically build evaluation metadata for grid services by using the information from the execution of different applications. A new MetaService AMDBS, automatic metadata building service, is designed to support this mechanism. The typical feature of this mechanism is that it can get not only the absolute evaluation information, but also the relative global ranking result through the collaboration of different AMDBSs. Web service group testing is used as an example of a grid application to accelerate the process of evaluation metadata building. Simulation experiment results show that the HP (heavy-powerful) algorithm which is designed based on the dynamic evaluation metadata can improve the application performance significantly, compared with the RR (round-robin) algorithm which does not use the evaluation metadata. The effect of using evaluation metadata in different task and resource conditions is also demonstrated by experiment results.