Virtual Insider

Blog archive

Shared-Nothing Live Migration: Cool, But Not a Game-Changer

Microsoft this week announced a new feature that will be available for Hyper-V version 3 and System Center Configuration Manager 2012: shared-nothing live migration. In so many words, it's the ability to move a virtual machine while it is powered on from one host to another without requiring shared storage.

While I am very excited to see this welcome feature for Hyper-V, which reinforces a much-needed notion of creativity at Microsoft, the feature is not a game changer by any stretch of the imagination. Let's take a look at how it works:

  1. Move the VM virtual disks, snapshots, vm configuration metadata over a standard Ethernet connection
  2. Move the VM state and memory
  3. Delete VM state on original host and complete live migrate to destination host

You can't deny that it is cool and might be appealing to a small subset of customer that are leveraging VMs with small virtual disks, but entertain this scenario for me: a VM with 1 TB virtual disk ... wait, Eli, you are not realistic, you might be thinking ... fair enough -- what about a VM with 500GB virtual disks? Moving that amount of data over a 1 GB Ethernet or even a 10 GB Ethernet is not quick or feasible in most environments, but assuming it was acceptable this would only be used as a maintenance technique which is what it was slated for anyway. To think that you will constantly be copying large virtual disks between hosts is not practical and is not scalable. How many VMs can you move at the same time? What is the impact on the network?

To top it all, you cannot use high availability with this feature, and that makes sense since you need a point of reference for HA to work. How can you recover a VM when its files are on the host that failed? Now maybe in the future they build in replication, but even then, constantly replicating is not easy, scalable or guaranteed. So, for those of you that were dismantling the SAN, not yet folks.

I do want to reiterate, however, that the feature is most certainly innovative and cool and adds value to Hyper-V 3, but I definitely expect VMware, Citrix and Red Hat to respond relatively quickly to such a feature. I am happy, though, that Microsoft is again leading in innovation and others are having to copy features, as opposed to Microsoft playing catch up all the time.

Your thoughts?

Posted by Elias Khnaser on 04/18/2012 at 2:56 PM


What is this?

Reader Comments:

Fri, Jul 20, 2012 ccRyder

Hasn't anyone here ever heard of VM6 Software's VMex product? VMex is a H/A Clustering software that doesn't use Shared Storage, nor does it use Microsoft's Clustering service. They have their own clustering service and it runs on Windows 2008 R2 Standard Server. They replicate all the data in real-time CDP. Check it out at http://www.vm6software.com Microsoft has tried but has not succeeded in replicating their product.

Fri, Jul 20, 2012

Doesn't anyone here eve heard of VM6 Software's VMex product ? VMex is a H/A Clustering software that doesn't use Shared Storage. They replicate all the data in real-time CDP. Check it out at http://www.vm6software.com Microsoft has tried but has no succeeded in replicating their product.

Wed, Apr 25, 2012 Chris Williams South Wales

Hyper-V will not take over from VMware, there is just not the support for High end infrastructures (DPM, Storage Load balancing/DRS, Storage IO etc. on multiple SANs, turning servers off for power saving) in Hyper-V... it's like they don't want to target the high-end market and VMware's pricin model gives the impression of a lack of interest in the smaller markets. I agree that Hyper-V is improving with this new revision but it isn't up to the standard set by VMware in version 4 and they have improved massively in version 5. There is a place for Hyper-V in smaller businesses (I speak as aa microsoft partner) but I'm afraid that's where it will always stay.

Wed, Apr 25, 2012 Chris Williams South Wales

This doesn’t provide HA and is only any good if you can move the VM files over a network (probably dedicated) quickly.

The best thing for them would be to leverage DFS-R (existing technology) to copy the VM files and keep them in sync on both servers (requirement for double disk space on both servers) then use that same technology to live migrate. That would provide HA and that would be a game changer. Why don’t I work for Microsoft again???

Thu, Apr 19, 2012 hypervfan

The VM is not moved before everything is verified over. So if there is a network disconnect during the move, the move fails, but the VM keeps running on the original host

Thu, Apr 19, 2012 Elias Khnaser Chicago, IL

@Belgium if this is about mobility, we already have that and it is more efficient with traditional Live Migration. Using Shared-nothing has challenges, my question is what happens if in the middle of a shared-nt-nothing migration a network is discnnected? will corruption occur? how do i recover when half the files are on one side and other half on another side? what happens if the initiating host has issues in the middle of a migration? at least with traditional live migrate, the file are statically located on the SAN, recovering is easier, but when the files are in flight it becomes complicated. NOW, AGAIN and for the RECORD, i am not saying it is not a cool feature or a welcome feature and i am happy Microsoft is innovating AGAIN and other are FOLLOWING. Damn i am using all the good keywords but BING can't seem to pickup on it. All that being said, however, this is NOT a GAME CHANGER by any strect of the imagination. is Hyper-v3 coming along and catching up? YES, will it close the gap with VMware? yes, will it over take VMware IN THE LONG RUN, YES, are we there now? NO will we be there in the next 3 years? NO. that is all i am saying. Eli

Thu, Apr 19, 2012 Elias Khnaser Chicago, IL

@anonymou a SAN that fails is a mis-configured, mis-architected and or mis-maintained SAN. @RW too much cool aid buddy :) no, i am just kidding, but buddy you are reciting to me all the features you are seeing at MMS, great, i am simply talking about shared-nothing as a feature and its relevance, not the entire conference or announcememnts. Eli

Thu, Apr 19, 2012 Elias Khnaser Chicago, IL

LMAO, vElias, i like that. Look, vElias criticizes VMware whne criticism is needed, i was one of the harshest critics of vRam when it came out and i have no problem saying a technology is not up to speed, for example i have written loads on VMware View and its limitations etc... but i am not here to defend myself or my track record, sorry Microsoft.To say that Shared-nothing is HUGE is drinking too much cool aid, mobility in the datacenter? please enlighten me how many VMs can you move at the same time over 1GB NIC? if i have a modest consolidation ratio of 10 - 1 and i want to evaucate my VMs from one host to another for maintenance, how long do i have to wait if the average sixe of a VM is around 100GB? and i am being nice by saying a 100GB, do you know how much traffic that will generate? i gave the feature a fair assessment but to say that it is HUGE is a stretch and when VMware and Citrix and other release this my position will be the same, this is not HUGE, it is simply a cool to have especially for small enterprise. vEli

Thu, Apr 19, 2012 Aidan Finn Ireland

vElias is at it again :) Seriously dude, just come out of your vCloset. Shared Nothing Live Migration is HUGE. To be able to have mobility in the datacenter between non-clustered hosts (important for hosters) and between clusters IS a big deal. I wonder how vElias will comment about the non-importance of this feature when VMware finally plays -catchup- :)

Thu, Apr 19, 2012 Aidan Finn Ireland

vElias is at it again :) Seriously dude, just come out of your vCloset. Shared Nothing Live Migration is HUGE. To be able to have mobility in the datacenter between non-clustered hosts (important for hosters) and between clusters IS a big deal. I wonder how vElias will comment about the non-importance of this feature when VMware finally plays -catchup- :)

Thu, Apr 19, 2012 Didier Van Hoye Belgium

this is not about HA. It's a VM mobility feature, even between clusters. You use it with or without HA in your environment. Whatever fits yoor purpose, need and budget. Mobility of VMs is very important, just like Live Migration. Whether or not you use HA here also depends on the environment. Please read up on this subject and the needs it serves in virtualization and a cloud infrastructure. It is a game changer as it allows for a great deal of no down time mobility scenarios where now you need to bring down the VM for longer or shorter periods.

Wed, Apr 18, 2012 RW Vegas

I'm at MMS for my company and it's been eye opening. Been taking furious notes all week and I think you're missing the point. It's not just shared nothing LM. It's shared nothing live migration, network virtualization, storage spaces, 32 proc vms, 1 tb memory WITHOUT vtax, Replica, the new vSwitch, PowerShell, SAN integration, 64 node clusters, bitlocker, new file server, new server manager, live storage migration, thin provisioning, deduplication, automated cluster updating (!!!), sr-iov, and I'm sure I missing more. Sounds like game changing to me.

Wed, Apr 18, 2012

I think it is a great idea for small businesses that want this ability. While you mention a host failing, you didn't mention that a SAN can fail also.

Wed, Apr 18, 2012 Elias Khnaser Chicago

LEM, sorry to tell you bud, HA can't work with this configuration. uf a host goes down the VM files become inaccessible, how are you supposed to bring the VM up on another host? Eli

Wed, Apr 18, 2012 JD

Lem, I'm intrigued; please explain how HA could function in this scenario where all the VM data & virtual disks reside on a failed server's local disks?

Wed, Apr 18, 2012 ordinaryitguy

What about Live Migration for VMs which has virtual HBAs and SAN LUNs attached? Also not a game changer VMW lover :) ?

Wed, Apr 18, 2012 Lem

You are wrong on one very important point, HA is perfectly usable in this configuration. The point is also not that you'd do it every day, but for specific needs the ability to evacuate an entire cluster while leaving the system running is useful. Denying its usefulness when it limits no scenarios is like denying the usefulness of storage migration in general.

Add Your Comment:

Your Name:(optional)
Your Email:(optional)
Your Location:(optional)
Comment:
Please type the letters/numbers you see above


Close this Advertisement