Essential elements for improved grid usability

A major obstacle to Grid usability today is that Grid applications are prone to failures and performance bottlenecks. Writing “better” Grid software cannot really solve this problem: to a certain extent, this an inherent property of a large and multi-layered system. Thus, troubleshooting is an essential Grid component. However, the very layers of middleware designed to shield Grid users from the variation in underlying systems can also hide the feedback needed to diagnose and troubleshoot poor performance or failure. This paper proposes two new Grid monitoring elements, Grid workflow identifiers and consistent component lifecycle events, that will make Grid failures easier to diagnose and troubleshoot.