question: benchmark details

[romange]

trafficstars

Hi,

how can we rerun the TCP benchmarks with PhotonLibOS ? Did you use client and server from https://github.com/alibaba/PhotonLibOS/tree/main/examples/rpc ?

Regarding the results in https://github.com/alibaba/PhotonLibOS#21-tcp Did you achieve 17.1Gb and 560K qps on a single core? Did I understand correctly?

[beef9999]

Wait me 5 minutes, I'm just about to upload the echo server code ...

[romange]

Thanks. Btw, very interesting project. Very glad to see that engineers start building high quality asynchronous frameworks for backend development. I developed mine https://github.com/romange/helio that have somewhat similar goals as PhotonLibOS but maybe a different design. I wonder if you gonna share HLD doc - maybe we could share ideas. Especially around using io_uring API efficiently.

[beef9999]

@romange I have updated the example https://github.com/alibaba/PhotonLibOS/blob/main/examples/perf/net-perf.cpp

In this test I started a coroutine to print QPS every second, but when CPU is 100%, the timing might lost accuracy. So one way is to increase the show_statistics_interval to large numbers, for example 10, to print the data once every 10 seconds. Another way is to observe the entire ingress/egress bandwidth of the test machine ( by iftop ), and calculate the QPS backwards.

Of course you will need to have two test servers.

[romange]

sure @beef9999, I sent you an email.

[beef9999]

haven’t received it…

[romange]

i've sent to beef9999 at qq. If you have not got it and it's not in your spam folder, it's quite disturbing.

can you please send me an email to roman at dragonflydb.io ?