added benchmark script

[zloyrusskiy]

No description provided.

[dizda]

Sorry why using proxychains here @zloyrusskiy ?

[zloyrusskiy]

I need a fast way to evaluate performance, I want to add serialization/deserialization of protocol stages for declarativeness and easier project support. Writing my own tool for benchmarks is a complex task, I didn't want to spend time on it. I thought iperf3 was a good option, but to make it work over socks5, you either need to compile a patched version or wrap it in a tricky way like this. But I looked at the overhead from this approach and realized that you are right. I'll try to explore this topic further.
For now, I'm looking towards:

https://pkolaczk.github.io/benchmarking-cassandra-with-rust-streams/

https://docs.rs/balter/0.4.0/balter/

https://github.com/tokio-rs/tokio/blob/e9ae5d4ce993d8c8611b86065f1fe26188fff0ca/stress-test/examples/simple_echo_tcp.rs

[dizda]

This is interesting, let me know how that goes

[dizda]

hey @zloyrusskiy

Any update on this? 🙏

[zloyrusskiy]

I'm working on it. I integrate with https://github.com/bheisler/criterion.rs

[dizda]

Thank you @zloyrusskiy

[dizda]

@valpackett wdyt about this bench?

[yuguorui]

In my humble opinion, perhaps performance testing should be done by cargo bench with handcrafted testcases, since it avoids introducing additional dependencies and extra performance overhead from the dependencies.

In addition, this performance test is basically testing the performance of tokio::io::copy_bidirectional.

[valpackett]

I guess it's better than nothing…?

this performance test is basically testing the performance of tokio::io::copy_bidirectional

That will come in handy for measuring the benefits of things like tokio-splice tho :)

[dizda]

Maybe having a docker conf to launch the bench script would be handy in this case, to avoid installing the libs locally on different architecture.