ReGS: user-level reliability in a grid environment

Grid environments are ideal for executing applications that require a huge amount of computational work, both due to the big number of tasks to execute and to the large amount of data to be analysed. Unfortunately, current tools may require that users deal themselves with corrupted outputs or early termination of tasks. This becomes incovenient as the number of parallel runs grows to easily exceed the thousands. ReGS is a user-level software designed to provide automatic detection and restart of corrupted or early terminated tasks. ReGS uses a web interface to allow the setup and control of grid execution, and provides automatic input data setup. ReGS allows the automatic detection of job dependencies, through the GRID-ADL task management language. Our results show that besides automatically and effectively managing a huge number of tasks in grid environments, ReGS is also a good monitoring tool to spot grid nodes pitfalls.

• Type: Conference
• Year: 2005
• Where: CCGRID
• Authors: J. A. L. Sanches, Patrícia Kayser Vargas, Inês de Castro Dutra, Vítor Santos Costa, Cláudio F. R. Geyer
• Keywords: Automatic Detection, CCGRID 2005, Cluster Computing, Early Terminated Tasks, Grid Environments