Problem with benchmarking virtual HDD

[marcelkg]

Hello guys

I’m benchmarking virtual machine in VMware workstation Pro and compare it with other hypervisors.

I have problem with benchmarking HDD efficient. I’am using cristal disk mark 6.0. I don’t know how but virtual machine windows 7 64 bit(guest OS) on wmware workstation pro 15.02 has better results than Host OS – Windows 10 pro. Is it possible?

Below are results from Host OS system.

Below are results from Guest OS system

These tests were repeated at least five times for every case: 500MB, 100 MB and 50MB. Presented tests are average results.

Can someone explain me what is happening? Maybe Vmware has some features what support and increase virtual disk efficient.

The same problem is in Passmark performance test Disk.

Below are results from Host OS system

HOST OS

DISK SEQUENTIAL READ – 180.5 MB/s

DISK SEQUENTIAL WRITE – 169.1 MB/s

DISK – RANDOM SEEK + RW - 9.7 MB/s

Below are results from Guest OS system

GUEST OS

DISK MARK 3471

DISK SEQUENTIAL READ – 358.6 MB/s

DISK SEQUENTIAL WRITE – 305.2 MB/s

DISK – RANDOM SEEK + RW – 296.0 MB/s

Virtual machine and HOST OS is placed on HDD disk SATA III.

Maybe I am doing something wrong. How should I do correct benchmark HDD and RAM?

CPU test looks correctly.

BEST REGARDS marcelkg

[gbohn]

Just a guess on my part, but maybe the file benchmark data is getting cached in the host memory.

If so, that might account for the better results (delayed physical writes and just re-reading the read data after the very first read caches it).

[marcelkg]

HD Tune TEST:

HOST OS

GUEST HOST

Does anyone has a idea how to correctly benchmark hdd virtual machine?

[marcelkg]

I made virutal machine windows 7 professional 64 bit SP1 in Virtualbox.

Result in virtualbox is about 32% worse than HOST OS. That result looks like is good.

Is there any correct method to benchmark vmware virtual machine hdd? Or maybe Vmware has some special features that improve efficient virtual hdd?

[marcelkg]

Does nobody can help?

[DashTheRunner1]

What I understand that we can't use the HDD benchmark tools on virtual disks, and allow me to explain why:

The virtual disk is actually one VMDK file that being handled by the kernel. So when you run the benchmark the HDD you actually reading and writing to one disk, and the IOPs are handled by the hypervisor/VMkernel.

I think the best way to monitor the virtual disks is by using ESXTOP.

Below are couple of articles to explain the functionality of ESXTOP and how to use it:

ESXTOP - Yellow Bricks

https://www.vladan.fr/storage-performance-troubleshooting-with-esxtop-guide/

I hope that helped.

Thanks,