Very interesting! Thank you!
These tests seem to be dealing with small messages. If the messages were larger the HTTP overhead might be negligible. I’d be interested in seeing the results with a larger payload involved.
A few ideas for more tests:
- Testing with HTTP/2. It has header compression and some other performance optimizations.
- Testing with GZIP or Brotli compression. A lot of servers are optimized to do compression and decompression very fast.
- Testing with GRPC (https://grpc.io/)
- Testing with Websockets