CS 790: A Java Universe for Condor
|
Abstract:
Condor is a distributed batch processing system which is a project at UW madison which utilizes the resources of idle workstations to help in executing computeintensive jobs such as simulations. Some of the main features of Condor include Remote process execution facility. Each machine belonging to the condor pool has a condor starter demon which can start a new job on its machine whenever it is asked to. Trapping I/O calls and redirect it back to the submitting machine so that local le system of the executing machine is not used. The use of this scheme will make condor run even in environments in which the submitting and executing machine are in dierent le domains and do not share a common le space like afs. The ability to checkpoint a executing process,so that it can be halted when its executing machine becomes non-idle and it can be transfered to a suitable idle machine.This involves writing out the state of the process to stable storage (usually a checkpoint le). This ensures progress of the work.
