FrameworkBenchmarks
FrameworkBenchmarks copied to clipboard
Some express js benchmarks are not using all CPU cores
OS (Please include kernel version)
(Code inspection)
Expected Behavior
Express js benchmarks should use all available CPU cores, or it should be documented why they are not.
Actual Behavior
Several express benchmarks only run on a single core, leading to a massive and obvious performance disparity between e.g. express-mysql and express-postgres (on fortunes: 44,166 RPS vs 718 RPS)
The core issue is that some of the express js tests have a cluster.fork code snippet that launches the express server on all CPUs, and some of them do not. Ones that don't are substantially underperforming. (express-postgres, express-graphql-*)
Steps to reproduce behavior
(Issue observed from public data on benchmark website)