TechEmpower Benchmarks

by Julian K. Arni

Recently some people have expressed dissatisfaction with the performance of Haskell web frameworks on the TechEmpower Framework Benchmarks.

I recently added servant to the benchmarks (though it will only feature in round 14). Rather than wait for the official results, I ran the benchmarks with a handful of frameworks on a dedicated server. The results are, I think, enough to put these concerns to rest.

Note that on some of the benchmarks, the spock code seems to have run across some issues (which I've not yet investigated it). Note also that some of the performance lead that servant has over snap and spock should probably be attributed to the fact that the servant benchmarks use the faster hasql database library. Finally, the results listed below only summarize the total number of requests (higher is better); for a more comprehensive picture, see the raw results