No abstract

No abstract

Achieving faster performance without increasing power and energy consumption for computing systems is an outstanding challenge. This paper develops a novel resource allocation scheme for memory-bound applications running on High-Performance Computing (HPC) clusters, aiming to improve application performance without breaching peak power constraints and total energy consumption. Our scheme estimates how the number of processor cores and CPU frequency setting affects the application performance. It then uses the estimate to provide additional compute nodes to memory-bound applications if it is profitable to do so. We implement and apply our algorithm to 12 representative benchmarks from the NAS parallel benchmark and HPC Challenge (HPCC) benchmark suites and evaluate it on a representative HPC cluster. Experimental results show that our approach can effectively mitigate memory contention to improve application performance, and it achieves this without significantly increasing the peak power and overall energy consumption. Our approach obtains on average 12.69% performance improvement over the default resource allocation strategy, but uses 7.06% less total power, which translates into 17.77% energy savings.

MPI libraries are widely used in applications of high performance computing. Yet, effective tuning of MPI colletives on large parallel systems is an outstanding challenge. This process often follows a trial-and-error approach and requires expert insights into the subtle interactions between software and the underlying hardware. This paper presents an empirical approach to choose and switch MPI communication algorithms at runtime to optimize the application performance. We achieve this by first modeling offline, through microbenchmarks, to find how the runtime parameters with different message sizes affect the choice of MPI communication algorithms. We then apply the knowledge to automatically optimize new unseen MPI programs. We evaluate our approach by applying it to NPB and HPCC benchmarks on a 384-node computer cluster of the Tianhe-2 supercomputer. Experimental results show that our approach achieves, on average, 22.7% (up to 40.7%) improvement over the default setting.

No abstract