Benchmarking GPRC in Rust and Go (medium.com)

It's curious that Rust performed as node did. I thought surely it would be better

It would be ideal to indicate what hardware and OS platform is being tested on. Ideally Linux and source code of your benchmark.

According to the reddit thread, the benchmarks were done on a pretty beefy MBP.

The macOS malloc is known to perform poorly, so I wonder if it's possible to narrow the gap just by using jemalloc, without any other optimisations.

  // Cargo.toml
  [dependencies]
  ...
  jemallocator = "0.3"

  // main.rs
  use jemallocator::Jemalloc;
  #[global_allocator]
  static GLOBAL: Jemalloc = Jemalloc;

If the issue is indeed repeated (de-)allocations, I would also like to see Linux glibc/musl benchmarks. I expect musl malloc to perform even worse, so many stories about allocation sensitive services performing horribly in production because they're based on alpine images.