HyperV is coming - Lookout VMWare

Posted by on October 27, 2008 under Uncategorized |

As I mentioned to some of my peers at the Round Table, I think Microsoft will eventually dominate the Virtualization market, and quickly start taking big chunks of VMWare’s customers away.  Here is some evidence: HyperV 2.0 Features Announced.

Now, if you know me, you know I am partial to FOSS (Free and Open Source Software - not necessarily free as in price), and Microsoft is not FOSS.  I tell you that as a disclaimer, so you know I am not a Microsoft fan trying to sell you on HyperV.  In IT, it’s a Microsoft world, so I have to study it, understand it, and use it.  I have been working with Virtualization in production environments for about 5 years now, most recently testing the viability of ESX and HyperV for our production use.

To the uninitiated, it is easy to pick HyperV.  I have already seen engineers, who would not otherwise consider virtualization, move to it ever since HyperV was made available as a part of the 2008 OS.  It makes sense.  With just a few simple clicks, suddenly you can run multiple OSes on your PC or server.  Dead simple.  It is also easy to use, and has some useful features already included.  The management console, snapshots, and import/export tools are of great value, and I would argue that it’s Quick Migration (as opposed to VMWare’s VMMotion or KVM’s Live Migration) is practical and meets almost all requirements. 

To the better informed, a closer look is needed, and that is what I have been doing for the past few weeks.  I can tell you straight away that HyperV suspiciously looks like Xen… a lot (With Windows 2008 as Dom0).  An interesting footnote: Microsoft’s longtime buddy Citrix purchased Xen Source last year.  Theoretically, that purchase would allow MS to obtain a private license with Citrix to use Xen code outside the open source license under which it was developed, and under a different license.  I can also tell you that side by side HyperV doesn’t match up as well with ESXi.  Even with similarly configured guests with identical host hardware, the tests I performed showed that ESXi was plain faster (on a side note, KVM was faster than both).  And the ESXi management features are impressive with the detail that you can manage your Guest VMs.  ESXi does binary translation, whereas HyperV requires virtualization extensions.  This means that ESXi will run on older hardware as well.

So if ESXi is technically better, than whats the deal with HyperV?  For evidence, see the above referenced article.  Microsoft will continue to work on HyperV, and will very quickly include all the features that the Enterprise requires (read that as all the features that VMWare has).  On top of that, it is included in their OS (akin to how Internet Explorer took over), and costs almost nothing to install it.  To the average IT person, this is a no-brain-er, and is a VERY easy to path to virtualization.  I can tell you that when I first started Virtualization 5 years ago, if HyperV existed and was as easy to install (and as functional) as it is today, we would be all HyperV by now.  That is saying a lot, coming from me.  And the performance difference, while noticeable, wasn’t so great as to exclude HyperV from consideration.

I hope VMWare can come up with a good way to compete with this juggernaut.  I think in time KVM will silently become the standard, especially in large production data centers, much like Apache and Linux are today.  But for other corporate uses, VMWare is going to have to complete heavily with HyperV.

For now, for us, KVM is our standard VM platform, and we will start using ESXi to fill in some holes (especially with older equipment).  By comparison, HyperV just doesn’t meet our requirements… yet.  But I have seen it already working well for others in production environments, and can only guess at what the future holds.