this value seems off, need to investigate whether its a bad thread output

We can output in csv format or json format to std out. or potentially output json to a file, each run could be line delimited.

Thoughts?

we may want to show which query params were slowest in the benchmark. could hint to use potential indexes we'd need.
Also since each query is used once and we are using non dedicated hardware (potentially). We can get skewed results running the benchmark a single time. We should run it many times times and averaging out the results. We can add p99 and p95s to remove long running outliers.

Architecture is documented here, updating with diagram.
https://github.com/shawnfeldman/timescale-benchmark/blob/main/docs/ARCHITECTURE.md

Would be great to discuss the implementation...

may take me a little bit to wire in gh actions

figure out how to populate the dataset in one step

we could utilize worker information produced in stats to show stats per worker.

We could show fastest and slowest workers potentially

median should be decimal when set size is even. using time.Duration is somewhat limited here

internal/benchmark/benchmark.go should have tests around stats collection

I have some basic sanitization done on headers and column sizes but have paused work there to focus on db queries and stats collection

Does this sound acceptable?

Will need to update the readme to add instructions to run it

Hopefully having time can just docker-compose the workflow for a demo and have a script to run the process locally