Open Stack

On Tue, 17 Nov 2015 11:14:29 +0100
Alexander Schmidt <alexs at linux.vnet.ibm.com> wrote:

> Hi all,
> 
> I started a blueprint [1] and spec [2] for enabling the usage
> of native AIO mode for disk devices. The idea is to enable it
> for storage backends/setups where IO performance benefits from
> using native AIO mode and where no downsides are known wrt
> stability or data integrity.
> 
> As there is a wide range of storage backends and setups, I'm
> looking for input on specific backends that are known to
> benefit from native AIO mode (or where known problems exist).
> These are the comments so far (copied from the spec):
> 
> * native AIO mode is a bad idea if the storage is not fully
>   pre-allocated, e.g. for qcow2 images that grow on
>   demand or sparse LVM storage
> * AIO mode has no effect if using the in-qemu
>   network clients (any disks that use <disk type='network'>).
>   It is only relevant if using the in-kernel network drivers
> 
> Cases where AIO mode is beneficial
> 
> * Raw images and pre-allocated images in qcow2 format
> * Cinder volumes that are located on iSCSI, NFS or FC devices.
> * Quobyte (reported by Silvan Kaiser)
> 
> Also input on the minimum libvirt/qemu version where native
> AIO mode should be used would be very helpful.
> 
> Thanks and regards,
> Alex
> 

Adding the links...

[1] https://blueprints.launchpad.net/nova/+spec/libvirt-aio-mode

[2] https://review.openstack.org/#/c/232514/