Predicting the performance of globus monitoring and discovery service (MDS-2) queries

Resource discovery and monitoring in a distributed grid environment gives rise to several issues, one of which is the provision of reliable performance and hence, the quality-of-service delivered to grid users. This performance requirement brings about the necessity to know how the monitoring and discovery service (MDS) would respond to various queries. We focus on the performance of the MDS as part of the knowledge needed by a grid entity, to choose a grid index information service (GIIS) for discovering resources. Several performance metrics are defined and the performance achieved by a GIIS is evaluated against that obtained by a GRIS [H.N. Lim Choi Keung et al. (2003)]. Based on the GIIS performance data collected, the values for the performance metrics are predicted using different algorithms. Thus, past performance observations can be used to qualitatively characterise the future performance of the GIIS, allowing grid middleware built on these services to be more predictable.