event loop runtimes benchmarks and mioco timeouts

[danoctavian]

You might want to check this out.

https://github.com/danoctavian/c10k-bench

I've gathered together some event driven runtimes and implemented an echo server for each.

For rust, i've put in mioco and coio.

The benchmark is crude and probably flawed. Also i'm not measuring latency average and standard deviation per request.

Any feedback is greatly welcome, i would like to make it a more solid benchmark, but i don't have much exp with benchmarks.

The trouble i am encountering with mioco is that requests time out when running 3000 connections simultaneously for 30s. if i run 128 threads each sequentially doing connections for 30s it completes.

The benching code is here:
https://github.com/danoctavian/c10k-bench/blob/master/bencher/src/Lib.hs

Thank you!

[dpc]

Hi. Can you try increasing slab size: https://github.com/dpc/mioco/blob/master/src/lib.rs#L1216 , as it limits the number of possible concurrent connections?

I'm very interested in tests like this, but I'm currently very short on time, so it might take me a week or two to reproduce anything.

[danoctavian]

yeah sure thing, i'll play around and let you know!

[dpc]

Hi, I've just pushed new version that implements multithreading, and the original problem you were hitting should be fixed too. It's working for me with go-bencher, I don't use haskell, so I haven't tried the haskell bencher.

[danoctavian]

just verified. now works without failing. Thanks!

[dpc]

Well, now you've got to update the scores. :P

[dpc]

I've created an issue in the original repo. Closing this one.