Re: [Qemu-devel] [PATCH v2 0/8] virtio-blk: multiqueue support

[Stefan Hajnoczi <stefanha@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 2835 bytes --]

On Mon, May 30, 2016 at 06:25:58PM -0700, Stefan Hajnoczi wrote:
> v2:
>  * Simplify s->rq live migration [Paolo]
>  * Use more efficient bitmap ops for batch notification [Paolo]
>  * Fix perf regression due to batch notify BH in wrong AioContext [Christian]
> 
> The virtio_blk guest driver has supported multiple virtqueues since Linux 3.17.
> This patch series adds multiple virtqueues to QEMU's virtio-blk emulated
> device.
> 
> Ming Lei sent patches previously but these were not merged.  This series
> implements virtio-blk multiqueue for QEMU from scratch since the codebase has
> changed.  Live migration support for s->rq was also missing from the previous
> series and has been added.
> 
> It's important to note that QEMU's block layer does not support multiqueue yet.
> Therefore virtio-blk device processes all virtqueues in the same AioContext
> (IOThread).  Further work is necessary to take advantage of multiqueue support
> in QEMU's block layer once it becomes available.
> 
> I will post performance results once they are ready.
> 
> Stefan Hajnoczi (8):
>   virtio-blk: use batch notify in non-dataplane case
>   virtio-blk: tell dataplane which vq to notify
>   virtio-blk: associate request with a virtqueue
>   virtio-blk: add VirtIOBlockConf->num_queues
>   virtio-blk: multiqueue batch notify
>   virtio-blk: live migrateion s->rq with multiqueue
>   virtio-blk: dataplane multiqueue support
>   virtio-blk: add num-queues device property
> 
>  hw/block/dataplane/virtio-blk.c |  68 +++++++++++----------
>  hw/block/dataplane/virtio-blk.h |   2 +-
>  hw/block/virtio-blk.c           | 129 +++++++++++++++++++++++++++++++++++-----
>  include/hw/virtio/virtio-blk.h  |   8 ++-
>  4 files changed, 159 insertions(+), 48 deletions(-)

There is a significant performance regression due to batch notify:

$ ./analyze.py runs/
Name                                   IOPS   Error
unpatched-d6550e9ed2             19269820.2 ± 1.36%
unpatched-d6550e9ed2-2           19567358.4 ± 2.42%
v2-batch-only-f27ed9a4d9         16252227.2 ± 6.09%
v2-no-dataplane                  14560225.4 ± 5.16%
v2-no-dataplane-2                14622535.6 ± 10.08%
v2-no-dataplane-3                13960670.8 ± 7.11%

unpatched-d6550e9ed2 is without this patch series.
v2-batch-only-f27ed9a4d9 is with Patch 1 only.  v2-no-dataplane is with
the patch series (dataplane is not enabled in any of these tests).

Next I will compare unpatched dataplane against patched dataplane.  I
want to make sure Patch 1 faithfully moved batch notify from dataplane
code to generic virtio-blk code without affecting performance.

If there is no difference then it means batch notify decreases
performance for some workloads (obviously not the same workload that
Ming Lei was running).

Stefan

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 473 bytes --]