Ring benchmark inspired by Programming Erlang: Software for a Concurrent World, by Joe Armstrong, Chapter 8.11.2
Write a ring benchmark. Create N processes in a ring. Send a message round the ring M times so that a total of N * M messages get sent. Time how long this takes for different values of N and M.
Be sure to do a release build.
$ cargo build --release
$ ./target/release/tokio-ring-bench 1000 1000I'm in the process of learning Rust, and have a relatively large tokio-based project I'm working on, and felt I needed some more experience
with tokio itself. This was done as a sort of drill or practice. I'm extremely comfortable with Akka, and I was curious how performance compared
across Akka, Actix, and raw tokio.
Currently, with tokio 0.1.18, results on my machine hover around .65 to .8 seconds.
This is roughly equal to the results I've seen for an Actix implementation of the same test, first surfaced here and then canonized here.
I saw times as low as .35 seconds for those runs, though lately it appears there's a performance regression in Actix and times are between 2 and 3 seconds.
I also implemented this in Akka (2.5.21) (TODO link to repo) and saw times averaging 0.6 seconds for the test. It's worth noting the Akka approach degrades as we increase actors and decrease roundtrips, but this tokio approach performs almost identically with inputs (100, 10000), (1000, 1000), and (10000, 100). In other words, using more tasks seems to add no overhead, but using more actors does, at least in the context of this silly benchmark.