Will Hyper-V 3 and SCVMM 2012 be as disruptive as people are saying?

Microsoft showed off a demo of Shared-Nothing Live Migration using Hyper-V 3 and SCVMM 2012, and twitter was awash with comments like @fordeg's "VMware got kicked in the nuts today at MMS 2012".

Yesterday during the Microsoft Management Summit keynote, Microsoft showed off a demo of Shared-Nothing Live Migration using Hyper-V 3 and SCVMM 2012, and twitter was awash with comments like @fordeg's "VMware got kicked in the nuts today at MMS 2012". And, while the technology certainly is cool and is something the industry has been waiting for, is it really all that disruptive?

For starters, it's unrealistic to think that VMware and Citrix aren't also working on solutions that provide similar functionality. Yes, Microsoft is ahead of the pack in that regard, but it's not like that was a secret feature that was dropped on us out of nowhere today. We've known about it since at least last September, possibly earlier than that. It stands to reason that both Citrix and VMware have something in the works, and we may learn of that functionality as early as Citrix Synergy in a few weeks. VMworld is in September, so we may have to wait that long to hear something from them.

So, in that regard, I'm not convinced that this will turn the platform virtualization world on it's head. Yes, it's cool. Yes, I'm excited to have this functionality. Ultimately, though, I expect to have this from every platform, and sooner rather than later.

What occurred to me later in the day was that the people cheering the loudest for this might be the companies that have aligned themselves with local storage. Companies like V3, Pivot 3 (something about that number 3), and even Dell with their R710 (part of the DVS reference architecture) stand to benefit the most from essentially free functionality. These hardware companies now get the benefit of the features without actually developing their own solution, like Nutanix did. Speaking of Nutanix, it's a good thing they have many other features, because this raises the level of competition between their product and the other appliance-based, local storage options.

Software companies also benefit. Citrix VDI-in-a-Box can immediately benefit from live migrations, since it was one of the main disadvantages of their local storage model. Other companies like Quest or Nexenta, who's built a solution that allows you to easily deploy VMware View on local storage and still use some advanced functionality like Linked Clones, can also benefit from this technology being built in to hypervisors. 

It's all pointing to the commoditization of hypervisors. Once everyone has the same functionality as the others and the playing field has been leveled, we can stop worrying about it from the desktop virtualization perspective and focus on other things. Platforms, management features, integration, and security all rise to the forefront. I can see a future where we don't get asked "which is better, XenServer, ESX, or Hyper-V?" 

What do you think? Is this kind of technology in Hyper-V disruptive, a non-starter, or the sign of the future where the focus is on other supporting technology? While we're at it, does it change your tone at all on the local versus SAN storage debate? It's certainly one more feature that you can take out of the SAN column.

Join the conversation


Send me notifications when other members comment.

Please create a username to comment.

While live migration without shared storage is cool. It doesn't allow for HA on server failure (loss of local storage). It will be interesting to see people use SMB 2.2 against hyper v clusters as a lower cost entry point.


For those interested in Hyper V3 features, feel free to download the: Windows Server “8” Beta Hyper-V Component Architecture Poster.



Haven't had hands on with it yet but what do the real requirements look like.  If I have a 100GB virtual file server running on local disk, will I need 10Gb to successfully migrate that machine?  What type of impact will that storage move have on other VMs at the source and destination?  How scalable is this so far as number of hosts and required network bandwidth, etc?  Lots of unknowns it seems.


Just wanted to mention that OpenVZ (the FOSS Linux containers aka OS Virtualization product that is upstream of Parallels Virtuozzo) has been able to do offine and online (live) migrations without shared storage since 2006... and I've been using it frequently since then.  Containers aren't what most folks are used to so it isn't exacly a perfect comparison but still.

Anyway, yeah... a vm will take longer to migrate because it has to transfer its filesystem from the source host to the destination... and then it will have to do a checkpoint and restore... and transfer that... but I would imagine that the network parameters could be tweeked such that if you wanted to reduce the network impact the transfer that's fine, it'll just take longer.  Or possibly you just have an additional nic dedicated to the "transfer network" that is separate from the outfacing network traffic.

To answer Gabe's question though... no... I don't think it is ground breaking or going to change much.  It will be seen as the "ghetto way" to migrate systems and will allow users with lower budgets to play when they previously couldn't... but I think SANs will still remain popular for some time.

VMware has added a number of local storage related features as well as general storage features as they have been moving away from a few of their storage partners... competing with their own partners... but I think everyone saw that coming anyway.

What is a mystery to me though is why shared storage has been required for so long to begin with.


rahvintzu: you can get pretty close to HA if you do a local Hyper-V Replica instead of putting it at a DR site. no automated failover, but smaller customers who don't have a SAN are generally not my first thought for needing automated HA.

I always love the call the first time someone has a torn write on a SQL server during a failure or testing and their automated HA just corrupts the database further.


@ Rich

Great point in regard to Hyper-V Replication, I fully agree on that approach to lower the outage window.