Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 4 Oct 2007 13:06:37 -0700 (PDT)
From:      Jeff Roberson <jroberson@chesapeake.net>
To:        Aryeh Friedman <aryeh.friedman@gmail.com>
Cc:        Artem Kuchin <matrix@itlegion.ru>, freebsd-current@freebsd.org
Subject:   Re: Scheduler selection for web hosting
Message-ID:  <20071004130501.A615@10.0.0.1>
In-Reply-To: <bef9a7920710041234i779891afp894941b571d63ed4@mail.gmail.com>
References:  <009a01c806bc$5c7021d0$0c00a8c0@Artem> <bef9a7920710041234i779891afp894941b571d63ed4@mail.gmail.com>

next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, 4 Oct 2007, Aryeh Friedman wrote:

> On 10/4/07, Artem Kuchin <matrix@itlegion.ru> wrote:
>> Hello!
>>
>> I have read that in 7-Current there are two schedulers.
>> 4BSD - which, AFAIK, is a renamed new SMP scheduler, but i'm not sure
>> ULE
>
> According to the scheduler team the only reason why ULE is not the
> standard scheduler is it has poor performance on single processor
> machines.

Well actually the only reason it isn't default is because we didn't ask 
for it until too late in the release cycle.  The poorer UP performance is 
at most 1-2% on real world tests.  It has about 10% higher context switch 
cost than 4BSD.  I have patches to improve this but they aren't quite 
stable yet.

Thanks,
Jeff

>
>>
>> 7-current amd64 is actually seems to be VERY stable on hardware and
>> software we use, so, we want to move it to production servers and
>> want to get max perfomance from it for web hosting.
>>
>> As, as i know, scheduler is a very important thing when i comes to
>> perfomance in havy loaded really multitasking system. We are having
>> about 900 processes in about 20 jails.
>>
>> So, what is the difference between the two? Which seems to be better
>> for hosting? Is ULE bugfree and stable enogh for this?
>
> ULE no question given your config... also from here on out I think it
> is the only one under active development and a high experimental
> version was tested last night and likelly to be further refined that
> should really fly on such enviroments.
> _______________________________________________
> freebsd-current@freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-current
> To unsubscribe, send any mail to "freebsd-current-unsubscribe@freebsd.org"
>



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20071004130501.A615>