Earlier tonight I ran h2load against our example Twisted server, serving its own source file, to compare it between CPython and PyPy, but also against nghttp2. This is because nghttp2 is a great comparison point: the closer we can get to its performance the better we're doing.
So let's have a website that compares them! Let's use something like codespeed to build a website that benchmarks all the HTTP/2 servers that I can find, in all their permutations. This would let us build up a corpus of results that we can aim to constantly improve over time. It'd also be really insightful, and probably would lead to some fun competition amongst different implementations.
@tatsuhiro-t @kazuho When I get this up and running I'll point you at it: you may find it fun! 😁
Earlier tonight I ran h2load against our example Twisted server, serving its own source file, to compare it between CPython and PyPy, but also against nghttp2. This is because nghttp2 is a great comparison point: the closer we can get to its performance the better we're doing.
So let's have a website that compares them! Let's use something like codespeed to build a website that benchmarks all the HTTP/2 servers that I can find, in all their permutations. This would let us build up a corpus of results that we can aim to constantly improve over time. It'd also be really insightful, and probably would lead to some fun competition amongst different implementations.
@tatsuhiro-t @kazuho When I get this up and running I'll point you at it: you may find it fun! 😁