[fn-mote]

The benchmark link doesn't work for me. (Blame my adblocker??) The graphics in the article are interesting but don't include a way to identify which are Rust/Java/Go/Python. Maybe you're supposed to assume they come in the order given, but the x-axis is time... could definitely be clearer. (Also apparently not from the OP, just the cited article.)

[jiehong]

Also, this kind of throughput in Java 11 in 2021 while Java 17 was already out is a bit meh.

Nowadays, with Java 21 and virtual threads, results might be quite different (and the default GC has changed as well).

I agree with the graphs being unreadable. The tables in the benchmark article are easier to read.

[ackfoobar]

Also, not using ZGC when latency is the focus just strikes me as unserious.

"Benchmarked on JDK 11, with G1"

[steveklabnik]

The first link is missing a "7" at the end, that's why it 404s. The second one works though: https://medium.com/star-gazers/benchmarking-low-level-i-o-c-...

"Rust, Go, Java, Python" in order, left to right, yes.

[mcronce]

There's an oddity in that article...the C++ p99.9 is lower than its p99. Am I crazy or should that not be possible?

[steveklabnik]

That is confusing to me as well, yes.

[mre]

Author here. The link is fixed thanks to M3t0r on GitHub: https://github.com/corrode/corrode.github.io/pull/6 The source code of the blog is open source. Contributions like these are very welcome.