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

Re: [Xen-devel] [PATCH] remove blocked time accounting from xen "clockchip"



>>> On 09.11.11 at 18:47, Laszlo Ersek <lersek@xxxxxxxxxx> wrote:
> On 11/09/11 14:35, Jan Beulich wrote:
>> That is, on an overcommitted system (and without your patch) I
>> would expect you to not see the (full) double idle increment for a not
>> fully idle and not fully loaded vCPU.
> 
> I tried to verify this with an experiment. Please examine if the 
> experiment is bogus or not.
> 
> On a four-PCPU host (hyperthreading off, RHEL-5.7+ hypervisor & dom0) I 
> started three virtual machines:
> 
> VM1: four VCPUs, four processes running a busy loop each, independently.
> VM2: ditto
> VM3: single VCPU running the attached program (which otherwise puts 1/2 
> load on a single CPU, virtual or physical.) OS is RHEL-6.1.
> 
> In VM3, I also ran this script:
> 
> $ grep cpu0 /proc/stat; sleep 20; grep cpu0 /proc/stat
> cpu0 10421 0 510 119943 608 0 1 122 0
> cpu0 11420 0 510 121942 608 0 1 126 0
> 
> The difference in the fourth numerical column is still 1999, even though 
> only 10 seconds of those 20 were spent idly.
> 
> Does the experiment miss the point (or do I), or does this disprove the 
> idea?

For one, my expectation may be wrong (while I think the consideration
of the accounting still being wrong even with the patch is correct).

Second, the amount of stolen time (presumably the second to last
column; not sure what kernel version RHEL-6.1 uses, so I can't
immediately verify) being just 4 is certainly too small to be relevant
(meaning that VM3's only vCPU got scheduled almost instantly in too
many cases, which I think is the intended behavior of the credit
scheduler in a contrived environment like this).

To get the amount of stolen time up, I think one would have to
penalize VM3 (so that it doesn't benefit from not having been
scheduled for 100ms each time). I don't, however, know how to
achieve that in practice.

One question certainly possible to answer is whether, with your patch
and across different (over-)load scenarios, process, system, idle and
steal times add up to wall time, which I don't think would be the case.
Which isn't to say that they do without your patch, just that it
addresses only part of a wider issue.

Jan


_______________________________________________
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®.