Job Steering Service in Interactive Grid Analysis Environment

Abstract

Traditional use of Grid computing has been to execute unattended batch jobs. In such batch oriented Grids, the grid user loses control over job as it is submitted and the only interaction between the job and the user is to get the status of the job, kill the job or to retrieve the final results. Interactive analysis of physics data requires an environment that provides the continuous feedback of the jobs. Physicists also need to retrieve the intermediate output of their jobs. The Steering Service in the interactive Grid-Enabled Analysis Environment (GAE) [1] is key element in establishing a useful interactive analysis system. This Service provides physicists an Application Programmer Interface (API) as well as Graphical user interface to get the continuous feedback of their jobs and allows them to control and steer the execution of the submitted jobs. Physicists are now able to move their jobs to different grid sites when needed. Acting autonomously, the Steering Service ensures the optimal execution of the jobs in Grid. Continuous monitoring of the job enables the Steering Service to resubmit the failed jobs hence providing the reliable execution of the job. Steering Service also ensure the optimal consumption of grid user’s resource quota by executing the job on the cheapest Grid site.