Instead of scaling the file count as node count increases, enable a mechanism to test a single file against all IO jobs, regardless of node count. This can fold into existing tests as an option, or be a net new test if that's very challenging or impossible.
Instead of scaling the file count as node count increases, enable a mechanism to test a single file against all IO jobs, regardless of node count. This can fold into existing tests as an option, or be a net new test if that's very challenging or impossible.