HPE Morpheus VM Essentials Software
1870510 Members
3436 Online
110549 Solutions
New Discussion

Re: VME: performance benchmark

 
vosgtu
Frequent Visitor

VME: performance benchmark

hello community,

I'm currently finalizing my POC to validate Morpheus Essential, and I've reached the performance testing phase. I'm seeing some fairly significant performance discrepancies in the MSSQL workload tests compared to VMware.

Basically, I deployed a Windows 2025 VM, installed MSSQL 2019 Standard, and the HammerDB tool.

The two infrastructures are 100% identical: compute, storage, network, etc.

I ran several tests to make comparisons:


- qcow2 vs. raw
- cache = none vs. cache = writeback

What I’ve noticed is that the server running on Morpheus never uses the VM’s cache and completely saturates the local disk.

As a result, the results are really poor compared to VMware, which manages the cache better and has lower disk latency.

Has anyone else experienced this?

 

4 REPLIES 4
MaartenV1
Occasional Advisor

Re: VME: performance benchmark

What VirtIO drivers version are u using?

There is a known issue with 0.1.302, you should try it with 0.1.271 if you are using 0.1.302

vosgtu
Frequent Visitor

Re: VME: performance benchmark

Hello 

thanks for the answer.

Currently, the driver is in version 0.1.266

i updated in 0.1.271 but didn't change the benchmark result

do you have some advice for a windows 2025 about the disk (qcow2 vs raw) and the cahce mode (none vs writeback) ?

 

 

CalvinZito
Neighborhood Moderator

Re: VME: performance benchmark

Thanks, that’s helpful. Since you tested both VirtIO 0.1.266 and 0.1.271 and the benchmark result did not change, I don’t think simply updating the VirtIO driver is the next thing to pursue.

The Morpheus documentation does recommend VirtIO drivers for optimal Windows VM I/O performance, including the appropriate VirtIO storage driver such as viostor. Source:  Infrastructure  ›  Clusters  › HVM Clusters › Guest Operating System Notes 

Since changing the VirtIO version did not affect the result, I’d instead focus on whether the two VMs being compared are actually equivalent from the guest through the storage path. In particular, I would compare:
  • The storage controller and driver actually in use by Windows, not just the installed VirtIO package version
  • VM vCPU and memory configuration
  • Disk type, size, and provisioning
  • The underlying storage/datastore used by each VM
  • The exact benchmark configuration, including block size, queue depth, read/write mix, test duration, and dataset size
That should help isolate whether the performance difference is coming from the guest configuration, virtual disk/controller configuration, underlying storage path, or the way the benchmark is being run.

While the eval license is intended as a "kick the tires" license and there isn't HPE support for your testing, depending on where you are, I might be able to get someone to take a look. Drop me a PM with details on who you work for and where you are. 


I work at HPE
HPE Support Center offers support for your HPE services and products when and how you need it. Get started with HPE Support Center today.
[Any personal opinions expressed are mine, and not official statements on behalf of Hewlett Packard Enterprise]
Accept or Kudo
CalvinZito
Neighborhood Moderator

Re: VME: performance benchmark

Also, and importantly, what version are you running. I assumed since you have the trial that you are on v9.1.0 but you could be on another version. What are you running? 



I work at HPE
HPE Support Center offers support for your HPE services and products when and how you need it. Get started with HPE Support Center today.
[Any personal opinions expressed are mine, and not official statements on behalf of Hewlett Packard Enterprise]
Accept or Kudo