Hacker News new | past | comments | ask | show | jobs | submit login

If you are trying to shave nanoseconds off an RPC you have architectural issues.



If you approach every program with that attitude you will never have an RPC subsystem where nanoseconds matter. You will be trapped in a self-fulfilling process where nanoseconds don't matter because the system is slow.


It's physically impossible for nanoseconds to matter for remote calls - most individual servers are larger than one light-nanosecond.


The efficiency of initiating the call limits how many calls the program can initiate per second, per thread. The potential latency of the response is irrelevant.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: