Sunday, October 21, 2018

Re: Benchmarking kernel, userland and Xenocara build processes

Hi again,

to elaborate a bit, the reason I'm interested in getting more detailed benchmark results for the build processes is that I've noticed there are many situations where servers with less hardware resources (CPU cores and RAM in particular) often seem to finish the builds quite a lot faster than ones with 4-8 times more cores and RAM.

I'm well aware that OpenBSD's SMP implementation is rather young, but I'd be curious to seeing the exact details. I have a hunch it's the disk I/O speed that's causing some of the build time differences of the servers.

Are there any "in-house" benchmarking tools the OpenBSD dev team is using to measure the build times?

- Jyri
--
+358-404-177133 (24/7)
jyri.hovila@turvamies.fi

No comments:

Post a Comment