Re: [Xen-users] Xen VMs and Unixbench: single vs multiple cpu behaviour

[Marko ÄukiÄ <marko.djukic@xxxxxxxxx>]

"Shell Scripts (8Âconcurrent)" is not slowed down, because even though one copy of Unixbench is runÂsimultaneously,Â

the test always uses all available CPUs (this is also mentioned in the USAGE file in the UnixBench package).

Regards,

Marko

On 24 November 2015 at 18:31, Dario Faggioli <dario.faggioli@xxxxxxxxxx> wrote:

On Fri, 2015-11-20 at 16:41 +0100, Marko ÄukiÄ wrote:
>Â
> 1 cpu VM (test A):
>
> Dhrystone 2 using register variables    Â3403.6
> Double-Precision Whetstone          785.5
> Execl Throughput                 Â1853.5
> File Copy 1024 bufsize 2000 maxblocks Â3909.4
> File Copy 256 bufsize 500 maxblocks   Â2468.3
> File Copy 4096 bufsize 8000 maxblocks Â6212.0
> Pipe Throughput                  Â2079.8
> Pipe-based Context Switching        Â1101.3
> Process Creation                  1811.4
> Shell Scripts (1 concurrent) Â Â Â Â Â Â Â Â Â Â3084.2
> Shell Scripts (8 concurrent) Â Â Â Â Â Â Â Â Â Â2838.6
> System Call Overhead             Â3511.4
> System Benchmarks Index Score     Â2407.4
>
> *******************
>
> 4 cpu VM:
>
> single copy of tests (test B-1):
>
> Dhrystone 2 using register variables    Â3355.0
> Double-Precision Whetstone          Â787.6
> Execl Throughput                  298.8
> File Copy 1024 bufsize 2000 maxblocks 3292.7
> File Copy 256 bufsize 500 maxblocks   2078.2
> File Copy 4096 bufsize 8000 maxblocks 5516.9
> Pipe Throughput                  1855.9
> Pipe-based Context Switching         999.9
> Process Creation                  Â254.4
> Shell Scripts (1 concurrent) Â Â Â Â Â Â Â Â Â Â 818.0
> Shell Scripts (8 concurrent) Â Â Â Â Â Â Â Â Â Â6493.1
> System Call Overhead              2870.2
> System Benchmarks Index Score     Â1564.2
>
> Process creation in test B-1 stands out in one more thing.
>

The reason why I asked for these results was to check whether it was
only one (or maybe just a few) specific benchmark(s), from the
UnixBench lot, that were dragging the final score down.

It turns out that, AFAICT, with the exception of "Shell Scripts (8
concurrent)", pretty much everything slow down.

Yes, "Process creation" is one of the unluckiest, but even "Execl
Throughput" or "Shell Scripts (1 concurrent)" take quite a hit.

> These are results for process creation of all 10 iterations:
>
> 1. 254,4ÂÂ
> 2. 266,0ÂÂ
> 3. 285,7ÂÂ
> 4. 439,2ÂÂ(!)
> 5. 272,3ÂÂ
> 6. 260,9ÂÂ
> 7. 279,4ÂÂ
> 8. 259,7
> 9. 265,1
> 10. 284,2
> As you can see, test number 4 stands out quite a bit.
>
Yeah, it confirms that something weird is up. However, even at its peak
value, it's still far fromÂ1811.4 (from the 1vCPU case).

Regards,
Dario
--
<<This happens because I choose it to happen!>> (Raistlin Majere)
-----------------------------------------------------------------
Dario Faggioli, Ph.D, http://about.me/dario.faggioli
Senior Software Engineer, Citrix Systems R&D Ltd., Cambridge (UK)

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