Workshop: 12th International Workshop on Runtime and Operating Systems for Supercomputers (ROSS)
Authors: Jan Bierbaum (Technical University Dresden); Maksym Planeta (Barkhausen Institute, Technical University Dresden); and Hermann Härtig (Technical University Dresden)
Abstract: Contemporary HPC systems run compute jobs on exclusively assigned hardware resources. During communication, polling for progress is used for minimal latency. Previous work on oversubscription and event-based communication shows these techniques can improve overall system utilization and reduce energy consumption. Despite these findings, neither of the techniques is commonly used in HPC systems today. We believe that the current lack of detailed studies of the low-level effects of event-based communication, a key enabler of efficient oversubscription, is a major obstacle to a wider adoption.
We demonstrate that the sched_yield system call, often used for oversubscription, is not best suited for this purpose on modern Linux systems. Furthermore, we incorporate event-based communication into Open MPI and evaluate the effects on latency and energy consumption using an MPI micro-benchmark. Our results indicate that event-base communication incurs significant latency overhead but also saves energy. Both effects grow with the imbalance of the application.