This PR does not yet add code for posting results to Conbench; that will be in the next PR. You can see where it will fit, though; we could post results as they're completed with run() or a finished set of results from a BenchmarkDataFrame.
I assume you'll also add in the cache dropping in the follow on PR as well so that we have similar results across the two runners? For now I would "just" copy that code over and implement it in arrowbench. Ideally it's part of some dependency-lite utility set that we could depend on — but I don't think we've yet to find a way to truly have an easy way to do that and it's 1(ish) bash command that copying around is the least-painful way to get to where we need to go.
Originally posted by @jonkeane in https://github.com/voltrondata-labs/arrowbench/pull/122#pullrequestreview-1233202215
Before switching running to direct from arrow-benchmarks-ci, we should copy over the code to clear the disk cache from voltrondata-labs/benchmarks.