Re: [Xen-devel] [PATCH 00/19] Assorted fixes and improvements to Credit2

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 17.06.16 at 19:32, <dario.faggioli@xxxxxxxxxx> wrote:
> |-------------------------------------|---------------------------------|
> | Xen build, high VM load, with noise | Iperf, high VM load, with noise |
> |-------------------------------------|---------------------------------|
> |            runq=core   runq=socket  |           runq=core runq=socket |
> | baseline     41.466       30.630    | baseline    20.333     20.633   |
> | patched      36.840       29.080    | patched     19.967     21.000   |
> |=======================================================================|
> 
> Which, summarizing, means:
>  * as far as Credit2 is concerned,  applying this series and using runq=socket
>    is what _ALWAYS_ provides the best results.

Always? What about the increase on far the right side of the above
table fragment? It's not a big change, but anyway.

> [*] Jan, I confirm that, with your series applied, I haven't yet seen any of
> those "Time went backwards?" printk from Credit2, as you sort of were
> expecting...

Well, that's better than I had expected then: I didn't really think
they would be gone entirely. How long of an uptime did your tests
cover? As noted in the cover letter, I've observed remaining odd
TSC/stime jumps to increase in range over time, with no explanation
so far.

Also I wonder whether I may translate your statement above to
a Tested-by for part or all of the series (right now there's only a
coding style fix to one of the patches and a slight extension to
the rdtsc_ordered() one pending for an eventual v2).

Jan


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