Linked by David Adams on Tue 16th Aug 2011 16:44 UTC, submitted by Michael
Post a Comment
You don`t really need this patch anymore. I can get 0.33ms latency with the regual kernel, with a good config. But it still outperforms the regular kernel, according to Con Kolivas benchmarks, so if you want the absolutely lowest latency, go for it.
Seems to perform at bit better aswell in other areas, according to this article !
Edited 2011-08-16 17:14 UTC
RE: Comment by graudeejs
by Neolander on Tue 16th Aug 2011 18:04 UTC
in reply to "Comment by graudeejs"
RE[2]: Comment by graudeejs
by Laurence on Tue 16th Aug 2011 19:10 UTC
in reply to "RE: Comment by graudeejs"
RE[3]: Comment by graudeejs
by Neolander on Tue 16th Aug 2011 20:09 UTC
in reply to "RE[2]: Comment by graudeejs"
RE[4]: Comment by graudeejs
by Laurence on Wed 17th Aug 2011 08:41 UTC
in reply to "RE[3]: Comment by graudeejs"
It is somewhat weird Linux doesn't ship with a set of schedulers to choose from since there is no scheduler suited for every possible workload. I think it might be a maintenance issue but I feel this could be an opportunity for some sort of plugin system for the kernel similar to drivers. There might be other areas as well like memory management where it could benefit.
RE: Comment by abstraction
by renox on Wed 17th Aug 2011 08:30 UTC
in reply to "Comment by abstraction"
It is somewhat weird Linux doesn't ship with a set of schedulers to choose from since there is no scheduler suited for every possible workload. I think it might be a maintenance issue
No, it's not (mainly) a maintenance issue, it's a policy choice: Linus wants to have a good "all-purpose" scheduler, not several special-purpose schedulers.. Of course, not everybody agree with him as shown by CK's BFS. Note that even if you don't use BFS, its existence had a nice side effect: it improved Linux, as it helped find a bug in the stock scheduler.




