Performance Evaluations in VMWare Workstation 12 Player

[SergeyKostrov]

*** Performance Evaluations in VMWare Workstation 12 Player ***

[SergeyKostrov]

[ Abstract ] Virtualization software, like WMware Player or Oracle VM Virtual Box, improves productivity and capabilities of software engineers by virtualizing some Operating System ( Guest OS ) inside another Operating System ( Host OS ). Virtualization software allows to reuse hardware to run several OSs at the same time. However, performance of processing is affected and results of testing clearly demonstrate it.

[SergeyKostrov]

[ Linux Ubuntu 16.04 LTS in VMware Player Workstation 12 on Dell Precision Mobile Workstation ] Intel Core i7-3840QM ( 2.80 GHz / Ivy Bridge / 4 cores ) Processor : Intel Core i7-3840QM Cores : 4 Processors ( CPUs ): 8 http://ark.intel.com/products/70846 Windows 7 SP1 is a Host OS. Linux Ubuntu 16.04 LTS is a Guest OS.

[SergeyKostrov]

[ Linux Ubuntu on IVB virtualized Test 1 ] ... ubuntu@ubuntu-vm:~/WorkEnvMIC$ ./hpctest1.out Starting Compute GFlops = 25.600, Secs = 4.039, GFlops per sec = 6.338 ... [ Linux Ubuntu on IVB virtualized Test 2 ] ... ubuntu@ubuntu-vm:~/WorkEnvMIC$ ./hpctest2.out Starting Compute on 2 threads GFlops = 51.200, Secs = 6.626, GFlops per sec = 7.727 ... [ Linux Ubuntu on IVB virtualized Test 3 ] ... ubuntu@ubuntu-vm:~/WorkEnvMIC$ ./hpctest3.out Starting Compute on 4 threads GFlops = 204.800, Secs = 14.818, GFlops per sec = 13.821

[SergeyKostrov]

[ Windows 7 on IVB real Test 1 ] ... ..\WorkEnvMIC>hpctest1.exe Starting Compute GFlops = 25.600, Secs = 3.233, GFlops per sec = 7.918 ... [ Windows 7 on IVB real Test 2 ] ... ..\WorkEnvMIC>hpctest2.exe Starting Compute on 2 threads GFlops = 51.200, Secs = 4.830, GFlops per sec = 10.600 ... [ Windows 7 on IVB real Test 3 ] ... ..\WorkEnvMIC>hpctest3.exe Starting Compute on 4 threads GFlops = 102.400, Secs = 6.365, GFlops per sec = 16.087 ...

[SergeyKostrov]

[ Summary ] Performance is affected when some processing is done in a virtualized environment: Test 1 is ~1.26x slower ( 7.918 / 6.338 ) Test 2 is ~1.37x slower ( 10.600 / 7.727 ) Test 3 is ~1.16x slower ( 16.087 / 13.821 )

[gaston-hillar]

Sergey,

Thanks for sharing this information. Can you share the test you are running? I mean, is it open source code?

[SergeyKostrov]

>>...Can you share the test you are running?.. Please use STREAM benchmark if you would like to evaluate performance of your system(s).

[gaston-hillar]

@Sergey,

Thanks for sharing the test you are running.

[SergeyKostrov]

I also recommend to read IDZ article Optimizing Memory Bandwidth in Knights Landing on Stream Triad at https://software.intel.com/en-us/articles/optimizing-memory-bandwidth-in-knights-landing-on-stream-triad

[gaston-hillar]

Sergey,

Thanks for the article recommendation. I had missed this one on IDZ, and therefore, your linked allow me to know about it and read it. :)