Re: video performance in rhel/centos 7 vs fedora 20

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 






On Tue, Jul 15, 2014 at 11:22 PM, Marc-André Lureau <mlureau@xxxxxxxxxx> wrote:
Hi

----- Original Message -----
> Hi All,
>
> We have been piloting deployments of VM using Fedora 20 host and guest
> and have been very happy with performance and stability in almost every
> respect, so a big thanks to all for this hard work. Things are looking
> great.
>
> Now we have been testing deployments of these same VMs (F20 guests) on
> rhel 7 (centos 7). We migrated one over and one of the issues seems to
> be a severe degradation of video performance vis-a-vis the F20 host.

The difference between rhel7 and f20 host is quite surprising. I get
quite bad performance with f20 host as well (spice 0.12.4-3, qemu 2.0.0-5)
​Hi, I got confused long time ago, actually, the performance ​between 
centos and fedora is quite different, as installing win7 guest for example, 
10min for centos host, but 1h for fedora host! 
Maybe the reason is the difference of IO behavior, but I'm not sure.
 
Are your host configured with the same spice settings (in particular
streaming-video and other compression settings) ?

> The setup is a C7 host, F20 guest and either C7 or F20 client, either
> local to the host or on the LAN.
>
> In this setup, the video plays fine for a few seconds, but eventually
> degrades to about 1fps with severe tearing of the video. Video does
> remain in sync with audio and there are no audio breakups.
>
> Any pointers on things to try?

I have been investigating some video rendering issues with rhel7 guest
on rhel7 host, see https://bugzilla.redhat.com/show_bug.cgi?id=1030024

However, even with the cogl fix and "PutImage BIG-REQUEST", the video
regions seems to vary too much for the Spice video filter to perform
well (using firefox/html5 video youtube). I believe it will need to be
improved.

(the cogl fix allows much more accurate region invalidation, the
big-request allows to update whole region atomically and not by chunks)

> I'm willing to rebuild some host components (e.g. spice-server or qemu
> or whatever) or try some patches or configuration changes to address
> this to track it down.

Your help would be most welcome to investigate and improve this further.
_______________________________________________
Spice-devel mailing list
Spice-devel@xxxxxxxxxxxxxxxxxxxxx
http://lists.freedesktop.org/mailman/listinfo/spice-devel

_______________________________________________
Spice-devel mailing list
Spice-devel@xxxxxxxxxxxxxxxxxxxxx
http://lists.freedesktop.org/mailman/listinfo/spice-devel

[Index of Archives]     [Linux ARM Kernel]     [Linux ARM]     [Linux Omap]     [Fedora ARM]     [IETF Annouce]     [Security]     [Bugtraq]     [Linux]     [Linux OMAP]     [Linux MIPS]     [ECOS]     [Asterisk Internet PBX]     [Linux API]     [Monitors]