Chapter 4

Benchmarking

Abstract

Benchmarking is a way to measure the performance of a supercomputer empirically. A benchmark provides some standardized type of workload that may vary in size or input dataset. Benchmarks provide as an output a metric that can be used for comparing the performance of different supercomputers based on that benchmark's specific workload. These performance metrics include floating-point operations per second, traversed edges per second, and giga updates per second. Because a supercomputer's performance can be compared and ranked based entirely on a single benchmark, multiple differing supercomputing performance rankings exist. A supercomputer that performs well using one specific benchmark may not perform very well when ...

Get High Performance Computing now with the O’Reilly learning platform.

O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.