ostmarco / actix-benchmark

Actix Benchmarks

Disclaimer

This benchmark was designed for a specific scenario. Do not take it into account when making decisions about your projects or your company's products.

If you are in any doubt, I suggest you do your own tests according to your scenario and using tools and methods that suit your case.

Abstract

The purpose of this benchmark is to test whether using Dynamic Dispatch, instead of Static Dispatch, exclusively on the dependencies of a Handler impacts the performance of an application considerably.

Benchmarks

Static Dispatch

$ wrk -t12 -c400 -d30s http://127.0.0.1:8080/primes

Running 30s test @ http://127.0.0.1:8080/primes
  12 threads and 400 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency   195.43ms   18.57ms 422.88ms   78.14%
    Req/Sec   102.81     39.24   280.00     66.58%
  36927 requests in 30.10s, 21.22GB read
  Socket errors: connect 155, read 242, write 0, timeout 0
Requests/sec:   1226.96
Transfer/sec:    722.02MB

Dynamic Dispatch

$ wrk -t12 -c400 -d30s http://127.0.0.1:8080/primes

Running 30s test @ http://127.0.0.1:8080/primes
  12 threads and 400 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency   197.47ms   21.05ms 423.41ms   78.12%
    Req/Sec   102.00     47.90   270.00     61.69%
  36515 requests in 30.09s, 20.98GB read
  Socket errors: connect 155, read 258, write 0, timeout 0
Requests/sec:   1213.40
Transfer/sec:    714.04MB

Conclusion

Although Dynamic Dispatch has a lower performance compared to Static Dispatch, this difference is minute and does not negatively impact the overall performance of the application.