Meltdown and Spectre impact on RDSH (SBC and XenServer 7.2) using Windows server 2012R2 and Server 2016
After all the news around Meltdown and Spectre it seems things are calming down now. Recent talks are focusing more on next-generation exploits rather than the performance impact. The Login VSI Meltdown and Spectre emergency edition has been a great success and helped many of our customers, vendors and community friends. I’d like to share some results from our internal performance validation labs where I’ve conducted tests on a large number of operating systems measuring the impact of patches on the OS, Hypervisor and BIOS level. Please keep in mind that this is a lab environment, results will vary on your own systems.
I focused on Windows Server 2012R2 and Windows Server 2016 for these tests and started with a clean image that was optimized for performance using VMware OSOT with every test I added a new layer of patches adding layers of mitigation one at a time.
|Clean||Windows Server 2012R2 (Feb 2018 Update),
Windows Server 2016 (Feb 2018 Update)
|Microsoft Patch (MS)||KB4056898 for Windows Server 2012R2,
KB4056890 for Windows Server 2016
|Microsoft & Hypervisor Patch (MS-HV)||Adding patch for XenServer 7.2* - XS72E015|
|Microsoft, Hypervisor and Microcode Patch (MS-HV-MCU)||Adding patch for Dell R730 BIOS ver. 2.7.1|
* As you can see the results in the blog post are surprising, for this reason we are going to test with XenServer 7.4 to see if the Meltdown and Spectre impact changes, stay tuned for the follow-up.
To be honest I expected the test results to be straightforward, however they are not entirely as I assumed before I started my research. Looking at the graphs it’s easy to see that the Clean test is by far the best performing for RDSH on Windows Server 2012R2, which is to be expected. However, when we look at the other 3 tests we see that once the Microsoft Operating System patch has been applied the performance does not degrade as significantly as we expected.
When we look at the results for Microsoft Windows Server 2016 we see that the patches overall have very limited impact confirming the claim that newer Operating Systems are hit less. The same also goes for more recent hardware, except for the Microcode patch, which seems to have the ‘biggest’ impact. But all in all, it does not seem to have a huge impact.
Of course as always, these tests are done in our lab and results may vary upon testing your own hardware and software combination.
VSImax Results (higher is better)
VSIBase results (lower is better)
Windows Server 2012 R2 % Impact (closest to 0 is best)
Windows Server 2016 % Impact (closest to 0 is best)
As was promised by Microsoft, the Microsoft patch itself seems to have sealed the leak on all levels of the Meltdown and Spectre issues. At least the performance impact seems to support this. As was told in the blogpost written by my colleague Blair we do see that the impact on 2016 is a lot less than we see on 2012R2.
The last few questions that I still have are:
- Does the performance impact changes when we change the patch sequence?
- Does the HV-MCU combination have the same effect as the Windows Patch, or are they all equal?
- How does optimization of the VM influence the test results?
These tests were performed on Citrix XenServer, however they have also been performed on VMware ESX, in my next post I will outline the results from those tests.
Automate, Automate, automate. In our VDILIKEAPRO test lab we have minimized human interaction (and thus error) by making sure all processes are executed fully automatic for example the base images are created using the Microsoft Deployment Toolkit where after PowerShell magic (thanks Henk & Sonny!) takes over and automatically clones the VM’s 6 times on a single host, prepares them for use, and runs 10 automated test cycles results are then averaged ignoring the first run.
All the connections to the target machines are done using RDP protocol. And the default workload, Knowledge worker workload, was used for all tests.
Important note: The results in the blogpost are surprising, for this reason will also test with XenServer 7.4 to see if the Meltdown and Spectre performance impact changes significantly changes as we have learned from tests with other hypervisors, stay tuned for the follow-up.