On Tue, Jul 27, 2010 at 10:52 PM, Adam Vande More <[email protected]>wrote:
> On Tue, Jul 27, 2010 at 2:29 PM, Debacker <[email protected]> wrote: > >> Of course, just like you could put real-time processes in one CPU, and >> normal processes on another to avoid implement complex algorithms. >> While your solution is pragmatic, I would like to know if there are >> "clean" >> ways to do it. If not, this would be a documented use case to why would >> anyone actually need an I/O scheduler. >> > > First, top-posting on this list is considered rude. Please don't do that. > Sorry, I didn't want to hurt anyone, I just didn't know the traditions of this mailing-list, I'll be careful from now on. If you're running 8.1, try "man gsched", it's new and haven't tried it. > Excellent! I can use 8.1, it's for a new setup. > Other than that, the traditional way would be to give higher priority to > the process that needs it. It's the poor man's io scheduler, but it > generally does work well. > > If you have lots of concurrent io and are running a UFS file-system, > consider running gjournal as it scales those requests better. > > Also if you're hardware supports it, NCQ is available via the ahci and a > few other modules. It will make your requests more efficient. > Thank you for your tips. I'm happy that I will be able to use FreeBSD for this job. Laurent Debacker _______________________________________________ [email protected] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[email protected]"
