All of lore.kernel.org
 help / color / mirror / Atom feed
* Equivalent of vmware SIOC (Storage IO Control)  in KVM
@ 2013-10-12  6:09 Soumendu Satapathy (sosatapa)
  2013-10-13 13:26 ` Paolo Bonzini
  0 siblings, 1 reply; 7+ messages in thread
From: Soumendu Satapathy (sosatapa) @ 2013-10-12  6:09 UTC (permalink / raw)
  To: kvm@vger.kernel.org, Stefan Hajnoczi

Do we have an equivalent of vmware SIOC like feature in KVM?

- soumendu

^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: Equivalent of vmware SIOC (Storage IO Control)  in KVM
  2013-10-12  6:09 Equivalent of vmware SIOC (Storage IO Control) in KVM Soumendu Satapathy (sosatapa)
@ 2013-10-13 13:26 ` Paolo Bonzini
  2013-10-13 22:13   ` Andrey Korolyov
  2013-10-14  2:18   ` Wangshen (Peter)
  0 siblings, 2 replies; 7+ messages in thread
From: Paolo Bonzini @ 2013-10-13 13:26 UTC (permalink / raw)
  To: Soumendu Satapathy (sosatapa); +Cc: kvm@vger.kernel.org, Stefan Hajnoczi

Il 12/10/2013 08:09, Soumendu Satapathy (sosatapa) ha scritto:
> Do we have an equivalent of vmware SIOC like feature in KVM?

Yes, you have two choices:

1) use cgroups to throttle I/O at the level of the host disk (i.e.
multiple virtual disks stored on the same disk share the limit).  If
you're using libvirt, add the <blkiotune> element to the definition of
the virtual machine.

2) enable I/O throttling in QEMU, to apply limits at the level of the
guest disk.  If you're using libvirt, add the "iotune" element within
the <disk> element in the definition of the virtual machine.

Both are documented at http://libvirt.org/formatdomain.html (search for
blkiotune and iotune).

Paolo

^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: Equivalent of vmware SIOC (Storage IO Control) in KVM
  2013-10-13 13:26 ` Paolo Bonzini
@ 2013-10-13 22:13   ` Andrey Korolyov
  2013-10-14  0:56     ` Fam Zheng
  2013-10-14  7:49     ` Paolo Bonzini
  2013-10-14  2:18   ` Wangshen (Peter)
  1 sibling, 2 replies; 7+ messages in thread
From: Andrey Korolyov @ 2013-10-13 22:13 UTC (permalink / raw)
  To: Paolo Bonzini
  Cc: Soumendu Satapathy (sosatapa), kvm@vger.kernel.org,
	Stefan Hajnoczi

Hello,

By the way, is there plans to enhance qemu I/O throttling to able to
swallow peaks or to apply various disciplines? Current one-second flat
discipline seemingly is not enough for uneven workloads especially
when there is no alternative like cgroups for nbd usage.

Thanks!

On Sun, Oct 13, 2013 at 5:26 PM, Paolo Bonzini <pbonzini@redhat.com> wrote:
> 2) enable I/O throttling in QEMU, to apply limits at the level of the
> guest disk.  If you're using libvirt, add the "iotune" element within
> the <disk> element in the definition of the virtual machine.

^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: Equivalent of vmware SIOC (Storage IO Control) in KVM
  2013-10-13 22:13   ` Andrey Korolyov
@ 2013-10-14  0:56     ` Fam Zheng
  2013-10-14  7:49     ` Paolo Bonzini
  1 sibling, 0 replies; 7+ messages in thread
From: Fam Zheng @ 2013-10-14  0:56 UTC (permalink / raw)
  To: Andrey Korolyov
  Cc: Paolo Bonzini, Soumendu Satapathy (sosatapa), kvm@vger.kernel.org,
	Stefan Hajnoczi

On Mon, 10/14 02:13, Andrey Korolyov wrote:
> Hello,
> 
> By the way, is there plans to enhance qemu I/O throttling to able to
> swallow peaks or to apply various disciplines? Current one-second flat
> discipline seemingly is not enough for uneven workloads especially
> when there is no alternative like cgroups for nbd usage.
> 

Hi,

In current upstream master there are improvements on throttling (actually a
total rework), which added burst limits with new options of bps_max, iops_max,
etc.

Fam

^ permalink raw reply	[flat|nested] 7+ messages in thread

* RE: Equivalent of vmware SIOC (Storage IO Control)  in KVM
  2013-10-13 13:26 ` Paolo Bonzini
  2013-10-13 22:13   ` Andrey Korolyov
@ 2013-10-14  2:18   ` Wangshen (Peter)
  2013-10-14  2:24     ` Fam Zheng
  1 sibling, 1 reply; 7+ messages in thread
From: Wangshen (Peter) @ 2013-10-14  2:18 UTC (permalink / raw)
  To: Paolo Bonzini; +Cc: kvm@vger.kernel.org, Jinbo (Justin)



on Sunday, October 13, 2013 9:27 PM, Paolo Bonzini wrote:
> Il 12/10/2013 08:09, Soumendu Satapathy (sosatapa) ha scritto:
> > Do we have an equivalent of vmware SIOC like feature in KVM?
> 
> Yes, you have two choices:
> 
> 1) use cgroups to throttle I/O at the level of the host disk (i.e.
> multiple virtual disks stored on the same disk share the limit).  If you're
> using libvirt, add the <blkiotune> element to the definition of the virtual
> machine.
> 
> 2) enable I/O throttling in QEMU, to apply limits at the level of the guest
> disk.  If you're using libvirt, add the "iotune" element within the <disk>
> element in the definition of the virtual machine.
> 
> Both are documented at http://libvirt.org/formatdomain.html (search for
> blkiotune and iotune).

Both "blkiotune" and "iotune" are only take effect on one same host.
How to throttle I/O across multiple host with shared storage devices?


^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: Equivalent of vmware SIOC (Storage IO Control)  in KVM
  2013-10-14  2:18   ` Wangshen (Peter)
@ 2013-10-14  2:24     ` Fam Zheng
  0 siblings, 0 replies; 7+ messages in thread
From: Fam Zheng @ 2013-10-14  2:24 UTC (permalink / raw)
  To: Wangshen (Peter); +Cc: Paolo Bonzini, kvm@vger.kernel.org, Jinbo (Justin)

On Mon, 10/14 02:18, Wangshen (Peter) wrote:
> 
> 
> on Sunday, October 13, 2013 9:27 PM, Paolo Bonzini wrote:
> > Il 12/10/2013 08:09, Soumendu Satapathy (sosatapa) ha scritto:
> > > Do we have an equivalent of vmware SIOC like feature in KVM?
> > 
> > Yes, you have two choices:
> > 
> > 1) use cgroups to throttle I/O at the level of the host disk (i.e.
> > multiple virtual disks stored on the same disk share the limit).  If you're
> > using libvirt, add the <blkiotune> element to the definition of the virtual
> > machine.
> > 
> > 2) enable I/O throttling in QEMU, to apply limits at the level of the guest
> > disk.  If you're using libvirt, add the "iotune" element within the <disk>
> > element in the definition of the virtual machine.
> > 
> > Both are documented at http://libvirt.org/formatdomain.html (search for
> > blkiotune and iotune).
> 
> Both "blkiotune" and "iotune" are only take effect on one same host.
> How to throttle I/O across multiple host with shared storage devices?
> 

Depending on how you use the shared storage, I think you could:

1) Throttle disk/network IO on the shared device with cgroups.

2) Set throttle limit (iotune) on each domain.

Thanks,

Fam

^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: Equivalent of vmware SIOC (Storage IO Control) in KVM
  2013-10-13 22:13   ` Andrey Korolyov
  2013-10-14  0:56     ` Fam Zheng
@ 2013-10-14  7:49     ` Paolo Bonzini
  1 sibling, 0 replies; 7+ messages in thread
From: Paolo Bonzini @ 2013-10-14  7:49 UTC (permalink / raw)
  To: Andrey Korolyov
  Cc: Soumendu Satapathy (sosatapa), kvm@vger.kernel.org,
	Stefan Hajnoczi

Il 14/10/2013 00:13, Andrey Korolyov ha scritto:
> Hello,
> 
> By the way, is there plans to enhance qemu I/O throttling to able to
> swallow peaks or to apply various disciplines? Current one-second flat
> discipline seemingly is not enough for uneven workloads especially
> when there is no alternative like cgroups for nbd usage.

As Fam said, QEMU 1.7 will support burst throttling.  I'm not sure what
the plans are for adding support to libvirt.

Paolo

^ permalink raw reply	[flat|nested] 7+ messages in thread

end of thread, other threads:[~2013-10-14  7:49 UTC | newest]

Thread overview: 7+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2013-10-12  6:09 Equivalent of vmware SIOC (Storage IO Control) in KVM Soumendu Satapathy (sosatapa)
2013-10-13 13:26 ` Paolo Bonzini
2013-10-13 22:13   ` Andrey Korolyov
2013-10-14  0:56     ` Fam Zheng
2013-10-14  7:49     ` Paolo Bonzini
2013-10-14  2:18   ` Wangshen (Peter)
2013-10-14  2:24     ` Fam Zheng

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.