It's always TCP_NODELAY. Every damn time. - Marc's Blog
brooker.co.za/blog/2024/05/09/nagle.html
What Jamie wrote
Distributed systems are hard to build and scale. There are so many components to making them work. And the depth of knowledge in all the various stacks makes it even harder. Including the network…
The first thing I check when debugging latency issues in distributed systems is whether TCP_NODELAY is enabled. And it's not just me. Every distributed system builder I know has lost hours to latency issues quickly fixed by enabling this simple socket option, suggesting that the default behavior is wrong, and perhaps that the whole concept is outmoded.
🤔