Re: [PATCH v3] migration/dirtyrate: make sample page count configurable

[Peter Xu <peterx@redhat.com>]

On Tue, Jun 01, 2021 at 08:09:34PM +0800, Hyman Huang wrote:
> Ping
> 
> though dirtyrate by sampling page may kind of be inaccurate,
> it still valuable for those who run qemu on non-x86 or kernel
> which does not support dirty ring, this patch is necessary i
> think, what would you think of it ?

Yes I think this patch is okay:

Reviewed-by: Peter Xu <peterx@redhat.com>

Maybe I can pick it up and repost with the hmp cmds as they conflict.

But note that even with this sample_pages parameter, my test still gets this
with a 200MB/s workload:

(qemu) calc_dirty_rate 10 16384
...
(qemu) info dirty_rate
Dirty rate: 21 (MB/s)

I think it means it does not solve the memory locality issue, so it may only be
useful for workload that mostly randomly distributed across all the ram.
However since normally this is used to evaluate "whether this customer VM can
be migrated", it also means maybe the admin has no idea about what type of
workload the guest is running.  Depending on this info will wrongly migrate a
very busy VM as the admin thought it's low loaded.

So I think at last if we want to make this feature to real use, we may need to
depend on either dirty logging or dirty ring to report the real numbers, even
without migration started.

-- 
Peter Xu