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

Re: [Xen-devel] Measure performance of hypercalls


  • To: "Dan Magenheimer" <dan.magenheimer@xxxxxxxxxx>
  • From: "George Dunlap" <George.Dunlap@xxxxxxxxxxxxx>
  • Date: Wed, 15 Oct 2008 16:43:58 +0100
  • Cc: xen-devel@xxxxxxxxxxxxxxxxxxx, Tommy Huang <nbp0204@xxxxxxxxx>
  • Delivery-date: Wed, 15 Oct 2008 08:44:21 -0700
  • Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=message-id:date:from:sender:to:subject:cc:in-reply-to:mime-version :content-type:content-transfer-encoding:content-disposition :references:x-google-sender-auth; b=p+uoLvAhG+UrKJFTWJTHUJd5msFEGfzUOA/WBF+o2RIHN+rcGUcrDluP9yUhJiGE/+ ED/g/Xr/LU2SAlGelSU9h0wzmjp1UunhbtcAvTptWAzsNg0bXNdWR3RVCoo55BlwNuDD +5Jzo5Jc1FLHw1+dBP+WUQjK9lpoIzRM3DuKQ=
  • List-id: Xen developer discussion <xen-devel.lists.xensource.com>

On Wed, Oct 15, 2008 at 3:00 PM, Dan Magenheimer
<dan.magenheimer@xxxxxxxxxx> wrote:
> Hmmm... lots of interesting things can happen between a hypercall
> entry and exit, eg. domain switch, interrupts, timer queues, etc.
> So just taking a pair of timestamps is not likely to have the
> desired result.

Domain switch you can factor in, if you trace runstate changes as well
(albeit a bit tricky).  Long-term averages are probably what you
really want (since a one-off isn't going to affect performance too
much).  So if you filter out hypercalls interrupted by a context
switch, and do an average over the rest, the interrupts, timer queues,
&c should more or less disappear in the noise (unless they're a
significant source of overhead, in which case they need some attention
anyway).

I have a tool that will do a bunch of this stuff for vmexit/enter
analysis.  If there were hypercall enter/exit timestamp traces, it
could be modified to do the same for hypercalls.  Then we might think
about fun things like analyzing hypercall continuations and what not.
:-)

 -George

>
> I too am interested in a similar measurement (for a specific hypercall),
> so if there IS a way to do this with reasonable accuracy, please
> correct me.
>
>> On Wed, Oct 15, 2008 at 3:12 AM, Tommy Huang
>> <nbp0204@xxxxxxxxx> wrote:
>> > Hi all,
>> >
>> > I am wondering how to measure the performance of hypercalls.
>> > For example, how many X86 instructions are executed in Xen
>> when you issue a
>> > hypercall.
>> > Also, how to get the frequency of hypercalls which is that how many
>> > hypercalls have been issued during a period of time?
>> > Thanks in advance.
>> >
>> > - Nb
>> >
>> > _______________________________________________
>> > Xen-devel mailing list
>> > Xen-devel@xxxxxxxxxxxxxxxxxxx
>> > http://lists.xensource.com/xen-devel
>> >
>> >
>>
>> _______________________________________________
>> Xen-devel mailing list
>> Xen-devel@xxxxxxxxxxxxxxxxxxx
>> http://lists.xensource.com/xen-devel
>>
>
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@xxxxxxxxxxxxxxxxxxx
> http://lists.xensource.com/xen-devel
>

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel


 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.