In this research project we are developing methods for predicting the that a job must wait in the queue prior to execution. These predictions will be used for co-scheduling of resources (such as computer servers, data servers, visualization mediums or remote instruments) in a distributed environment. Currently, we have developed methods for predicting application run time based upon historical informations. Our methods have been shown to be up to 64% better than current methods.
This research was supported by Argonne National Laboratory with funds from DARPA for the GLOBUS project. The funds supported 1 Ph.D. student. This work has generated 4 conference papers and 1 journal submission.