[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Xen-devel] Re: performance of credit2 on hybrid workload

  • To: George Dunlap <george.dunlap@xxxxxxxxxx>, xen-devel@xxxxxxxxxxxxxxxxxxx
  • From: David Xu <davidxu06@xxxxxxxxx>
  • Date: Tue, 31 May 2011 20:55:16 -0400
  • Cc:
  • Delivery-date: Tue, 31 May 2011 17:56:42 -0700
  • Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=mime-version:in-reply-to:references:date:message-id:subject:from:to :content-type; b=mRvhocG8wqC6NgdrD05av13gYf5xPwQLH81hpcKBEpVjDFi/sKuLrMlGfWVIJAjTw8 spvZgNd+b/HUzVvSoWDlNRA6jPJgcpqAo+vap2uOjqkGJmY6ynID0UqCE2sdkbwAU6uQ Ax5EIxIWUuDj+oie0xzLy2RwlrayUWQdtAj3M=
  • List-id: Xen developer discussion <xen-devel.lists.xensource.com>


I want to reduce the latency of a specific VM. How should I do based on credit scheduler? For example, I will add another parameter latency besides weight and cap, and schedule the vcpu whose VM holds the least latency firstly each time. Thanks.


2011/5/26 George Dunlap <george.dunlap@xxxxxxxxxx>
Please reply to the list. :-)

Also, this is a question about credit1, so it should arguably be a
different thread.


On Wed, 2011-05-25 at 19:34 +0100, David Xu wrote:
> Thanks. The boost mechanism in credit can significantly reduce the
> scheduling latency for pure I/O workload. Since the minimum interval
> of credit scheduling is 10ms, the magnitude of latency for the target
> VM should be 10ms (except the credit is not used up and vcpu remain
> the head of runqueue ) as well. Why the real latency in my test (Ping
> the target VM) is much shorter than 10ms? Does the vcpu of target VM
> remain the head of runqueue if it was boosted?
> David
> 2011/5/25 George Dunlap <george.dunlap@xxxxxxxxxx>
>         On Mon, 2011-05-23 at 09:15 +0100, David Xu wrote:
>         > Hi,
>         >
>         >
>         > Xen4.1 datasheet tells that credit2 scheduler is designed
>         for latency
>         > sensitive workloads. Does it have some improvement on the
>         hybrid
>         > workload including both the cpu-bound and latency-sensitive
>         i/o work?
>         > For example, if a VM runs a cpu-bound task burning the cpu
>         and a
>         > i/o-bound (latency-sensitive) task simultaneously, will the
>         latency be
>         > guaranteed? And how?
>         At the moment, the "mixed workload" problem, where a single VM
>         does both
>         cpu-intensive and latency-sensitive* workloads, has not been
>         addressed
>         yet.  I have some ideas, but I haven't implemented them yet.
>         * i/o-bound is not the same as latency sensitive.  They
>         obviously go
>         together frequently, but I would make a distinction between
>         them.  For
>         example, an scp (copy over ssh) can easily become cpu-bound if
>         there is
>         competition for the cpu -- but it is nonetheless latency
>         sensitive.  (I
>         guess to put it another way, a workload which is
>         latency-sensitive may
>         become i/o-bound if its scheduling latency is too high.)
>          -George

Xen-devel mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.