search

andrewthad / rotera

Persistent rotating queue
BSD 3-Clause "New" or "Revised" License
8 stars 1 forks source link

Criterion benchmarks #7

Open chessai opened 5 years ago

chessai commented 5 years ago

With graphs

chessai commented 5 years ago

Not sure what data we should benchmark.

chessai commented 5 years ago

@andrewthad

I think a good place to start would be benchmarking push without committing, and push with committing, and push with intermittent committing

What do you think? Any other ideas?

andrewthad commented 5 years ago

The only way I can imagine doing benchmarks with criterion would be to make the mmap backing the database MAP_ANONYMOUS and then to just insert a ton a data. Committing would be a no-op so the frequency of commits would be irrelevant.

A real benchmark, where data is actually persistented, would be trickier to do. I don't think that criterion, which is intended to be used for microbenchmarks, would be suitable for this.

chessai commented 5 years ago

What would be suitable? Rolling something on our own?

andrewthad commented 5 years ago

Yes, we could still have it running on a single host, but we would just clock a giant write of around a gigabyte of data.