Re: [Xen-devel] Something about xenalyze

[George Dunlap <George.Dunlap@xxxxxxxxxxxxx>]

On Fri, Dec 14, 2012 at 8:01 PM, Qi Zhang <cheungrck@xxxxxxxxx> wrote:

I am a Ph.D. student, and get to something about xenalyze to ask.

Part of my output of 'xenalyze --scatterplot-runstate' is:

1v0 10.822094482 2
1v0 10.822094482 3
0v0 10.822096386 3
0v0 10.822096386 0

It seems like that VCPU0 of VM1 has two states(2 and 3) at the same time point(10.822094482). So, can you help me to explain that?

Thank you!

The granularity that xentrace reports is only down to the nanosecond; things that happen less than a nanosecond apart may appear to happen at the same time.  However, even with a 4GHz clock speed, 1ns is still only 4 clock cycles (if I'm doing my math right) -- it seems unlikely that both trace records could be taken in that time.

Did you see any warnings about tsc skew?

If you're really keen to track this down, you'd have to do "xenalyze --dump-raw --scatterplot-runstate", so that we could see the raw TSC values for those records, and on what pcpu each ran.

 -George

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel