Hi all. I known that i can shape disk iops using cgroups. What about internal qemu shaper ? http://wiki.qemu.org/Features/DiskIOLimits does it work with xen? (i want to use it under xen 4.3) -- Vasiliy Tolstov, e-mail: v.tolstov@selfip.ru jabber: vase@selfip.ru
On Fri, 6 Dec 2013, Vasiliy Tolstov wrote:> Hi all. I known that i can shape disk iops using cgroups. What about > internal qemu shaper ? http://wiki.qemu.org/Features/DiskIOLimits > does it work with xen? (i want to use it under xen 4.3)I haven''t tried it myself but given that it works on the underlying storage, I think it might work with qdisk on Xen. (qdisk is just the interface exposed to the guest and it is completely decoupled from the way the image is stored.)
2013/12/8 Stefano Stabellini <stefano.stabellini@eu.citrix.com>:> I haven''t tried it myself but given that it works on the underlying > storage, I think it might work with qdisk on Xen. (qdisk is just the > interface exposed to the guest and it is completely decoupled from the > way the image is stored.)Thanks! Is that possible that this can works with phy:/ device? (i''m use lvm) -- Vasiliy Tolstov, e-mail: v.tolstov@selfip.ru jabber: vase@selfip.ru
On Mon, 9 Dec 2013, Vasiliy Tolstov wrote:> 2013/12/8 Stefano Stabellini <stefano.stabellini@eu.citrix.com>: > > I haven''t tried it myself but given that it works on the underlying > > storage, I think it might work with qdisk on Xen. (qdisk is just the > > interface exposed to the guest and it is completely decoupled from the > > way the image is stored.) > > > Thanks! Is that possible that this can works with phy:/ device? (i''m use lvm)phy devices use the in-kernel blkback as default block backend. You would need to specify backendtype=qdisk to use QEMU and qdisk instead. However keep in mind that QEMU is slower than the in-kernel blkback.