In a new head-to-head comparison published by independent technology assessment firm Principled Technologies, VMware vSphere once again trounces challenger Hyper-V by delivering superior speed, performance, and reliability for zero-downtime virtual machine migrations. Live VM migration is crucial for proactive maintenance on hypervisor hosts and for distributing workloads as demand shifts — optimizing for either performance or power savings. Once an exclusive feature of the industry-leading vSphere platform, live migration has become yet another casualty of the checkbox war — where a simple “yes” or “no” cannot sufficiently convey vital technological differences.
Microsoft may position Hyper-V Live Migration as good enough, but a side-by-side comparison clearly reveals that an imitation is never as good as the original. VMware vMotion continues to improve and vSphere 5 enjoys many new enhancements, including support for multiple 10GbE interfaces to increase bandwidth for migration traffic — complementing the well-proven ability to migrate up to 8 VMs at a time.
Hyper-V Live Migration, introduced two years ago after much delay, will evidently remain stagnant for quite some time to come. In fact, the shipping version of Hyper-V can still only accommodate a single migration at a time — whether a source or a destination. This leads to the somewhat disingenuous claim that Hyper-V supports up to 8 concurrent Live Migrations per [16-node] cluster!
Principled Technologies conducted migration testing in two different scenarios, providing objectivity to correct the inaccurate claims of parity between the platforms. In the first scenario, one host in a cluster is running 10 VMs and is put into maintenance mode. Comparing elapsed times to evacuate each hypervisor quantifies migration speed without conjecture. In the second scenario, a single busy VM is moved from one host to another. The elapsed time is considered, but more important is the performance impact to the application undergoing migration.
Take a look at the results and see for yourself:
Host Evacuation: Entering Maintenance Mode
Under real-world conditions, VMware vMotion is over five times faster than Live Migration. When Patch Tuesday rolls around and it’s time to update Windows-based hypervisors, be sure to build plenty of margin into the maintenance window to allow for host evacuations. What’s amazing about this victory is that although vSphere 5 now supports multiple 10GbE NICs for vMotion traffic, this scenario used just a single interface on each host.
Tier-One Application: Degradation Caused by Migration
A busy database VM is expected to deliver continuously; minimally impacting application performance during migration is a crucial requirement when demanding SLAs must be met. VMware vMotion cranks out 63 percent more orders than Hyper-V during a 4-minute migration window. In fact, vSphere needed just half a minute to move a very busy 4-vCPU SQL Server VM with 16GB RAM. The charts below make it clear: vSphere empowers IT architects to virtualize with confidence.
Reliability: The Prime Directive
Having the best possible performance is not the only thing that matters — live migration would be of limited use if it caused workloads become unstable or crash. Therefore, imagine the surprise to learn that Hyper-V VMs fail in a reproducible, albeit unpredictable, way with the dreaded Blue Screen of Death (BSOD) after undergoing Live Migration. I’m sure this will be eventually resolved through some future Patch Tuesday, but that’s not what consumers of your IT infrastructure want to hear.
Hyper-V: “Missing Something” or “Good Enough”?
Last, but not least, an enthusiastic Hyper-V advocate and long-time VCritical reader once provided this revealing insight:
Thanks to Principled Technologies, we now lay this Microsoft conjecture to rest.
VMware vSphere: The Best Performance and Reliability
The success of a cloud deployment depends entirely on the foundational infrastructure. Not only does VMware vSphere provide the best performance density per host, it also offers the best operational management capabilities for your cloud. Don’t fall for the “good enough” trap!
Be sure to download the full report from Principled Technologies.
still reading the report, but your statment “Live VM migration is crucial for proactive maintenance on hypervisor hosts” is mostly incorrect it would be reactive maintance not proactive. Proactive means there has been planning involved, yes vmotion and live migration allows reactive migrations and seemingly from this report vmware trumps microsoft. In my experience this doesn’t seem quite true, yes I have both enivro’s running on the exact same hardware. Also personally I’ve never seen a stability issue, yes I’ve heard about them but in almost all cases there was a hotfix to cover the issue.
just had to stir the pot
Another Hyper-Victory for our ESXi champion !
Always a nice time to read your posts Eric, thanks.
Glad to have you as a reader!
well now you know why vmware does not allow bad results to be published. Yes, there are flaws in the testers work but they are flaws that are designed into windows thus I do fault microsoft and not the testers. I would never run w2k8r2 built in nic drivers for a 10gb pipe. The difference is night and day as far as performance is concerned.
And now you will be able to have true enterprise virtualization for free…
http://www.ovirt.org/
Download will be available after the workshop.
http://www.ovirt.org/news-and-events/workshop/
So what did you do “plan” it at the last second? Doesn’t sound like much of a plan, usually when we need to evac a host we know about with more then enough time to evac. Now if I knew a wrecking ball was about to hit one of my servers and I was only given 30 minutes then yes a fast vmotion would be nice but I’ve never seen that type of situation before.
I try to keep a open mind being a customer to either hyper-v or Vmware…
And i question a report that in the first graf chooses to emphesize size of the VMs. As fare as i know live migraiton needs shared storage and so size doesn’t matter when we talk live migration. Size of the VMs memory does..
So basicly you are giving me data from something that can’t be done…
I see this as yet another VMware propaganda attempt, and i know MS isn’t any better
Casper,
Thanks for your comment. The report does not compare VM storage, any references to size would be memory.
If you would like to review all of the technical details, please see the full report. It is very transparent.
Eric
Thanks for your quick reply.. well as you may know i would rather compared it to server 2012 and hyper-v 3.0 rather then Hyper-v 2.0… 🙂
I haven’t read your report but i take it its hyper-v 2.0.. and not hyper-v 3.0
Even if the document is 8 month old, i affraid windows server 2012 and hyper-v 3 is not GA yet. Microsoft is not a time travelling agency yet 😉
NiTRo that was somewhat my point though it may not have been clear..
I know its a bit late i see this article, but from my current perspektive im more considering Hyper-v 3.0 compared to Vmware ESXi..
Its allso a fare better competing product then Hyper-v 2.0
Looking forward to seeing any comparisons..