High performance computing (HPC) scheduling landscape currently faces new challenges due to the changes in the workload. Previously, HPC centers were dominated by tightly coupled MPI jobs. HPC workloads increasingly include high-throughput, data-intensive, and stream-processing applications. As a consequence, workloads are becoming more diverse at both application and job levels, posing new challenges to classical HPC schedulers. There is a need to understand the current HPC workloads and their evolution to facilitate informed future scheduling research and enable efficient scheduling in future HPC systems. In this paper, we present a methodology to characterize workloads and assess their heterogeneity, at a particular time period and its evolution over time. We apply this methodology to the workloads of three systems (Hopper, Edison, and Carver) at the National Energy Research Scientific Computing Center (NERSC). We present the resulting characterization of jobs, queues, heterogeneity, and performance that includes detailed information of a year of workload (2014) and evolution through the systems’ lifetime (2010–2014).

The global address space (GAS) programming model provides important potential productivity advantages over traditional parallel programming models. Languages using the GAS model currently have insufficient support from existing performance analysis tools, due in part to their implementation complexity. We have designed the Global Address Space Performance (GASP) tool interface that is flexible enough to support instrumentation of any GAS programming model implementation, while simultaneously allowing existing performance analysis tools to leverage their tool's infrastructure and quickly add support for programming languages and libraries using the GAS model. To evaluate the effectiveness of this interface, the tracing and profiling overhead of a preliminary Berkeley UPC GASP implementation is measured and found to be within the acceptable range.