A Fault Avoidance Strategy Improving the Reliability of the EGI Production Grid Infrastructure

摘要

Reliability is a crucial issue for the development of stable and effective production grid infrastructures. That is, grid users must be able to trust upon the runtime service they request and receive from the underlying grid. Many runtime services and capabilities offered by modern Grid infrastructures are not available in advance to the application developers and dynamically bound only at the execution time, leading to an increased incidence of interaction faults. In this work we propose, implement and evaluate a novel low-impact fault-avoidance scheme, specifically conceived to improve the grid reliability from the user/application point of view, by providing proper service status information to the workload management system. In particular, starting from the EGEE experience, we designed a strategy inhibiting the use of some specific runtime capabilities on the available resources as soon as the monitoring system detect any anomalous behavior associated to these capabilities and re-integrating them when they restart to correctly work again. The results of a significant set of tests ran on the production EGEE infrastructure, have been presented to show the effectiveness of our approach.