Optimizing Repetitive Resource Accesses with Stateful Web Services

Many modern scientific applications consist of numerous executions of legacy programs. The repetitive but discontinuous resource access pattern in these applications causes challenge in solving them with grid technologies. A stateful Web service model is proposed, which exploits the factory/instance pattern to schedule resources according to job submitters instead of submitted jobs. Each instance represents a customized execution environment that is exclusively accessed by one job submitter, and can be accessed repetitively to perform as many jobs as desired. At the same time, this model introduces an individualizing operation for every instance, which provides a mechanism to minimize the amount cost of remote data accesses required by all jobs submitted to the instance. This service model is exploited in enabling and optimizing repetitive resource accesses that occur in AS applications