Multi-site scheduling with multiple job reservations and forecasting methods

Most previous research on job scheduling for multi-site distributed systems does not take into consideration behavioral trends when applying a scheduling method. In this paper, we address the scheduling of parallel jobs in a multi-site environment, where each site has a homogeneous cluster of nondedicated processors where users submit jobs to be executed locally, while at the same time, external parallel jobs are submitted to a meta-scheduler. We use collected load data to model the performance trends that each site exhibits in order to predict load values via time-series analysis and then perform scheduling based on the predicted values.