Resource Sharing Usage Aware Resource Selection Policies for Backfilling Strategies

Job scheduling policies for HPC centers have been extensively studied in the last few years, specially backfilling based policies. Almost all of these studies have been done using simulation tools. All the existent simulators use the runtime (either estimated or real) provided in the workload as a basis of their simulations. In our previous work we analyzed the impact on system performance of considering the resource sharing of running jobs including a new resource model in the Alvio simulator. In this paper we present two new Resource Selection Policies that have been designed using the conclusions reached in our preliminary work. First, the Find Less Consume Distribution that attempts to minimize the job runtime penalty that an allocated job will experience. Based on the utilization status of the shared resources in current scheduling outcome and job resource requirements, the LessConsume policy allocates each job process to the free allocations in which the job is expected to experience the lowest penalties. Second, we have also described the Find Less Consume Threshold Distribution selection policy which finds an allocation for the job that satisfies the condition that the estimated job runtime penalty factor is lower than a given value Threshold.