[Spice-devel] Unfair comparisons with RDP

Sat Jul 2 06:30:05 PDT 2011

----- Original Message -----
> From: "Alon Levy" <alevy at redhat.com>
> To: "John A. Sullivan III" <jsullivan at opensourcedevel.com>
> Cc: spice-devel at lists.freedesktop.org
> Sent: Saturday, July 2, 2011 1:43:44 PM
> Subject: Re: [Spice-devel] Unfair comparisons with RDP
> On Sat, Jul 02, 2011 at 07:24:59AM -0400, John A. Sullivan III wrote:
> > On Sat, 2011-07-02 at 12:27 +0200, Alon Levy wrote:
> > > On Sat, Jul 02, 2011 at 05:45:34AM -0400, John A. Sullivan III
> > > wrote:
> > > > On Sat, 2011-07-02 at 04:56 +0200, Alon Levy wrote:
> > > > > On Fri, Jul 01, 2011 at 09:40:41PM -0400, John A. Sullivan III
> > > > > wrote:
> > > > > > On Fri, 2011-07-01 at 17:05 +0200, Alon Levy wrote:
> > > > > > > On Fri, Jul 01, 2011 at 03:00:32PM +0200, Gianluca Cecchi
> > > > > > > wrote:
> > > > > > > > On Fri, Jul 1, 2011 at 1:04 PM, John A. Sullivan III
> > > > > > > > wrote:
> > > > > > > > > Interesting observation. That is true; we did not
> > > > > > > > > create separate VM
> > > > > > > > > definitions for SPICE and TSPlus thus the TSPlus
> > > > > > > > > environment is using
> > > > > > > > > the QXL driver. Would we expect that to have any
> > > > > > > > > "supercharging" effect
> > > > > > > > > on RDP?
> > > > > > > > >
> > > > > > > > >
> > > > > > > >
> > > > > > > > Probably not, because afaik (that is not so much ;-)
> > > > > > > > Remote Desktop
> > > > > > > > (and probably tsplus too) works at the GDI call level,
> > > > > > > > so it should
> > > > > > > > not depend so much on video adapter/video driver...
> > > > > > > > It was simply a question that arose analysing how to
> > > > > > > > correctly
> > > > > > > > replicate comparisons...
> > > > > > > > Coming back to the test case and these operations:
> > > > > > > >
> > > > > > > > rdp
> > > > > > > > 17: display desktop, i.e., minimize all open
> > > > > > > > applications
> > > > > > > > 42: Paint existing LibreOffice document, i.e., restore
> > > > > > > > from minimize
> > > > > > > >
> > > > > > > > spice
> > > > > > > > 61: display desktop, i.e., minimize all open
> > > > > > > > applications
> > > > > > > > 92: Paint existing LibreOffice document, i.e., restore
> > > > > > > > from minimize
> > > > > > > >
> > > > > > > > I think they are GDI ones, so that naturally when using
> > > > > > > > rdp they are
> > > > > > > > executed locally on client desktop (only the gdi
> > > > > > > > directives are sent),
> > > > > > > > while in spice (raster?) they will be network intensive
> > > > > > > > (from a slow
> > > > > > > spice implements a driver, which implements a large part
> > > > > > > of the gdi api. So any
> > > > > > > operation that it doesn't implement is done via the
> > > > > > > windows gdi software rendering
> > > > > > > and the result given to the driver (which is spice) as an
> > > > > > > image.
> > > > > > >
> > > > > > > So in cases where the specific operations are all
> > > > > > > implemented by the driver the
> > > > > > > performance should be identical. In other cases spice will
> > > > > > > be suboptimal, since
> > > > > > > it will send the image and not the operation. In both
> > > > > > > cases the rendering should
> > > > > > > be correct.
> > > > > > >
> > > > > > > > link point of view).
> > > > > > > > So probably an optimized rdp could never be beaten on
> > > > > > > > too slow links?
> > > > > > > >
> > > > > > > ><snip>
> > > > > > Hmm . . . I remember you saying that the Windows product was
> > > > > > actually
> > > > > > more developed than the Linux product. Could it be that you
> > > > > > have
> > > > > True
> > > > >
> > > > > > implemented more of the GDI API than the X API (or whatever
> > > > > > one uses for
> > > > > > Linux) and thus my Linux client is more regularly falling
> > > > > > back to
> > > > > > sending images rather than directives?
> > > > > Client != Guest. A confusion that arises all the time here :)
> > > > > The client
> > > > > is *using* the graphics api on whatever platform. The linux
> > > > > client uses
> > > > > pixman mostly. The windows client uses gdi. The gdi canvas (as
> > > > > the graphics
> > > > > backend for the clients is called) has seen more usage /
> > > > > optimization I think,
> > > > > so you are not wrong in your conclusion. There are actually
> > > > > two different clients
> > > > > right now, spicec and any client based on the spice-gtk, such
> > > > > as vinagre or spicy.
> > > > > Could you try any of the later to see if you get 100% cpu with
> > > > > them as well?
> > > > <snip>
> > > > Sorry - I realize I stated that backwards! However the 100% CPU
> > > > problem
> > > > is a different one. We are noticing that the Windows server
> > > > viewed via
> > > > the Debian client is laggard but CPU utilization is fine on both
> > > > client
> > > > and server. The problem with 100% CPU utilization is when we
> > > > have a
> > > > Fedora 15 server.
> > > by server you mean guest? So this is the driver taking 100% cpu?
> > >
> > <snip>
> > Yes, I hope I have my terminology right. Host is the system running
> > KVM, server is the system running on the KVM host, and client is the
> > device I am using to see the server by connecting to the host :) If
> > there is a more official terminology, do please correct me as the
> > right
> > vocabulary seems to be one of the most difficult things to master in
> > SPICE <grin>
> >
> The terminology is:
> host - machine running vm processes
> spice server - part of the vm process.
> guest - whatever is running in the vm.
> qxl driver - the part of spice running in the guest
> spice client - spice viewer, possibly on another machine
> 
> So I was trying to understand if you mean that you were running top on
> the host
> and seeing the process take 100% cpu, or running top inside the guest
> and seeing
> X (we are talking about a F15 guest, right?) take 100%. The later
> suggests a driver
> problem, while the former a server problem (or just a guest doing a
> bloody lot of work).

As I understood the problem observed by John - John has F15 guest with KDE4 (I was close to think It was specific for KDE4) and he is observing 100% CPU usage caused by Xorg process on that guest. But That implies High CPU usage on host of qemu process as well indeed. I personally didn't see such high cpu usage - 100% but I can say It is quite high (60-90%) even doing very easy work on guest. I even filed a bug to track that https://bugs.freedesktop.org/show_bug.cgi?id=38522 (sorry about not mentioning), there are some profiles from sysprof which John kindly provided. So you can take a look:). Soren promised he would take a look at that as I believed It was qxl driver issue but I dunno about progress on that issue and yes I believe client (spicec/spicy) has no impact on that.

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