Microsoft Hyper-V: the day after

June 26 Microsoft released its first bare-metal hypervisor: Hyper-V 1.0.

Obviously, when a market leader like Microsoft releases a new key product like a hypervisor in an exploding industry like virtualization, and re-focuses every department on it, there are some repercussions.

It doesn’t really matter if the product is long overdue, if it comes with a limited feature-set, or if it the market is already crowded. The fact that Microsoft is (finally) fully committed on virtualization has repercussions on the landscape, at strategical and financial levels.

What follows is a summary of the first reactions from the market players with some additional analysis from us.


The partners and competitors positions

Citrix is currently the stronger partner for Microsoft despite it recently acquired a product, XenServer, which can compete with Hyper-V on every aspect (they even have a similar architecture) and in many cases has more features.

The two companies already declared their cooperation strategy in the past, so Citrix had not much to say but wait for some of our products able to interact with and manage Hyper-V (the reference is to XenDesktop 2.0).

Now that Microsoft product is out we’ll see if this synergy is really positive to Citrix (like the years-old one on Terminal Services) of self-defeating.

The Virtual Iron position is very different.
The company offers a Xen-based hypervisor like Citrix and it’s currently targeting the SMB market, exactly the one that Hyper-V can attract at this first release.

Virtual Iron tries to knock down the Microsoft product using four arguments: the new hypervisor obliges to adopt a new OS (Windows Server 2008), it’s a 1.0 release so customers won’t trust it in production environments, it lacks of many features that most customers want, Microsoft is not the best vendor to run mixed data centers where some virtual machines are powered by Linux.

Most of these statements are arguable as follows.
There’s no difference in adopting Windows Server 2008 (when it includes the hypervisor) or adopting another virtualization platform because in this special case the new OS is just acting as an engine. Customers are free to run Windows Server 2003 (or older) virtual machines pretty much like on Virtual Iron.
While it’s true that Hyper-V lacks of many features it’s not true that all of them are indispensable. A fairly large number of small companies are still at a point in their adoption of virtualization where the only need is to achieve a deadly simple server consolidation. And Hyper-V can achieve that task.
It’s hard to believe that a Microsoft-based hypervisor can’t virtualize well a Linux virtual machine (even if Microsoft made the huge mistake to only support Novell OS at RTM date), while a Xen-based hypervisor instead can perfectly virtualize a Windows virtual machine. And considering that 95% of the virtualized systems are Windows it’s easy to image what is the customers’ priority.

But Virtual Iron is right on one thing: at today there’s a well-known distrust in Microsoft products at their first release. 
Avoid the deployment of a 1.0 product in production environment is considered a savvy approach by many IT managers, no matter which vendor they are considering, but in Microsoft case there is a special tradition in refusing the first company attempt in any enterprise area.
Microsoft will have to work hard and release Hyper-V 2.0 as soon as possible to change the negative mind set that the market has against it.

Sun is much more moderated in its answer (no link available) but it’s expected considering the interoperability agreement in place with Microsoft.

Pretty much like Citrix, Sun is building a complete virtualization portfolio, including a bare-metal hypervisor (xVM Server), a connection broker (VDI) and a cross-platform management solution (Ops Center), which will imply a full competition with Microsoft.
Despite that, exactly like Citrix, Sun welcomes Hyper-V and confirms its commit to support it on its products.

At the same time anyway, Sun is glad to underline what are seen as major shortcomings in the new product: the lack of live migration will leave customers in the cold for unplanned outages, Hyper-V has unknown chance to compete with open source competitive products because of the faster implementation of new features, Hyper-V can only support Windows guest OS and this doesn’t aid the power consumption reduction, Hyper-V can’t scale up, Hyper-V can’t manage physical environments.

Like for Virtual Iron almost every statement is arguable and it’s surprising how weak the Sun position is trying to discourage the adoption of new product.
The lack of live migration certainly is a key feature that some enterprise customers need, but it seems pretty evident that Microsoft is not trying to address that audience with this first version of the hypervisor.
This also address the second point: a SMB market is less concerned about the last cutting-edge feature as long as the basic requirement of server consolidation is already satisfied.
Hyper-V doesn’t support Windows guest OSes only, but even in that case the large majority of virtualized systems are Windows, so the hypervisor seems to address most of the customers need and seems to help enough in the power consumption reduction effort.
Maybe the architecture that Microsoft chosen for Hyper-V doesn’t scale up as well as the one used by VMware ESX (and this must be verified in any case), but Sun is using exactly the same one on its upcoming xVM Server.
Last but not least the fact that Hyper-V can’t manage physical and virtual infrastructures is totally irrelevant: like Sun has its Ops Center so Microsoft has System Center Operation Manager (SCOM) to accomplish the task. It’s not a duty for a hypervisor.

Parallels has a much easier approach (no link available).
The company is about to release a server virtualization product (and it’s unclear if it will be able to compete with Hyper-V and other bare-metal hypervisors or not), but it’s presence primarily depends on the OS virtualization technology Virtuozzo Containers, widely adopted in the hosting market.

Because of this Parallels simply states that Hyper-V doesn’t represent a threat to them, but doesn’t miss the opportunity to highlight that the product doesn’t support most Linux, BSD and Solaris guests.

Obviously the VMware‘s position (the very long version) is the easiest to guess.
Being the market leader the company is the one that can lose the most if Hyper-V takes too much credibility.

To the company Hyper-V is not reliable enough, is not secure enough, has not a small footprint, has a too limited feature-set, has not a broad portfolio of product which integrate with it.

Even in this case, some of these statements are arguable.
Despite what VMware says the Hyper-V footprint is actually smaller than the ESX one (less than 1MB against 32MB) but the difference in the architectural approach makes very c
ompl
ex to really compare the two.
Microsoft has a broad range of products which will integrate and support Hyper-V soon: the obvious System Center Virtual Machine Manager (SCVMM), Operation Manager (SCOM), Configuration Manager (SCCM), Data Protection Manager (DPM) and more coming. Together these products extend the capabilities offered by the hypervisor alone and may satisfy a large part of the SMB market.
It’s true that Microsoft is not currently able to offer all the virtualization-based solutions that a customers may want (namely the connection broker for VDI environments and the virtual lab management solution for semi-automated development and testing environments). But it must be seen if a SMB is interested in any of these applications. 
The well-known Microsoft strategy in any area where it invests it’s to let its partners explore the market opportunity and then proceed acquiring them or competing with them (pretty much what VMware did in the VDI and P2V migration space so far).

But VMware is right on at least a couple of points.
First of all the Hyper-V reliability must be proven (the migration of MSDN and TechNet web sites on it is not enough).
Also the security level of the product must be proven, but the choice to use a full version of Windows for the Parent Partition (which is fully trusted by the hypervisor) greatly extends the Hyper-V attack surface (no matter if the Server Manager can turn down all services that are not used by the hypervisor).
Additionally, the Hyper-V kernel is not the Windows kernel (which has a trustable level of maturity) but a new thing, which Microsoft never detailed too much.
In any case both Microsoft and VMware will have to handle with code vulnerabilities in the same way: no vendor can assure that its code is more secure than another one.

 

The stock market reaction

VMW_hyperV

CTXS_hyperV

MSFT_hyperV

 

The analysts perspective

Not only the the entrance of Microsoft in the virtualization has repercussions. It also creates confusion among customers. And the industry analysts are urged to give their advice. 


Chris Wolf
, Senior Analyst at Burton Group, is well balanced in its analysis: Hyper-V lacks of some key feature (live migration, memory over-commit) but the product is stable and comes for free, and this is an acceptable trade-off for a 1.0 version.
Microsoft couldn’t wait any longer to seriously enter the market and had to sacrifice the higher end of its planned feature set.

Wolf even says that Hyper-V may be considered for production use at its first release because of the cost, the first motivating factor for many customers. For this reason he expects a price drop for ESX.

And if the price of the hypervisor falls the competition will move on the management layer.
There Microsoft has a competitive advantage with a suite which can seamless manage physical and virtual infrastructures, while VMware is still fully focused on virtual machines only.
Therefore the market leader will have to close solid partnerships with enterprise management vendors like HP, IBM, CA and BMC to stay ahead of Microsoft.

Brian Madden, independent analyst at BrianMadden.com, is much more neat in his predictions.
He sees Hyper-V deeply influencing the relationship between Microsoft and Citrix, at a point where the latter drops Xen as the underlying virtualization engine and adopts Hyper-V by 2009-2010.

While this scenario seems pretty unlikely at beginning, it has some reasons to exist: the move would bring balance back in the relationship, giving Citrix the opportunity to innovate on top of a Microsoft engine (like always happened for Terminal Server) instead of having a full replication of whole stack.

To justify this major change, Madden says that the current market penetration of XenServer (as is, powered by Xen) is near zero. And since it’s very unlikely that Microsoft may have the same kind of acceptance, Citrix may prefer to jump on that bandwagon rather than driving its own train.

To give further solidity to its thesis Madden highlights that the Citrix business model is not made to deal with the the open source community and that the company never had much commitment on it so far.

So when Citrix will drop Xen the community around it will disintegrate, moving en mass to KVM and deeply impacting on those other companies that rely on the open source engine: Virtual Iron, Oracle (which may move to KVM much earlier because of Red Hat), Novell and Sun.

 

The new landscape

The key point is not what features Hyper-V has or has not, how many VMs can host in a highly dense environment, but which kind of audience it’s trying to serve.

As virtualization.info wrote in April 2007, VMware left uncovered the whole SMB market, totally focusing its flagship product around the needs of the biggest enterprises.
This granted the company the highest chance to survive even after the Microsoft entrance in the market and maybe this was the original place since the early days.

Microsoft knew, because of its history, that Hyper-V couldn’t approach the enterprise market until at least its second release, so when obliged to decide where to put its effort, the company decided to concentrate on providing a stable and fast virtualization platform, with the minimum amount of features to help the largest part of the market in the fist step: the plain and simple server consolidation.

The adoption of virtualization technologies is still so limited that there’s no need to compete on the high-level features from day one. The big profit is at SMB level, where Microsoft knows how to move better than anyone else.

So it’s very likely that VMware will start to re-adjust its approach as soon as ESX 4.0 is released, exclusively offering what’s beyond server consolidation: automation first, cloud computing then.
If so, at that point the real competitor for VMware will not be Microsoft but Citrix, which is already working to build value beyond server consolidation.

What the ecosystem will do?

It’s clear that at the moment Microsoft is the best partner to work with: Hyper-V leaves plenty of room for enhancements by 3rd party companies, and Microsoft may even decide to buy a few to consolidate its position.
VMware instead is becoming a much less comfortable partner to deal with: in the effort to stay ahead of competition and build the dynamic data center that we all hope to have, the company is investing in all areas, giving few chances to build value on top of its platform (the VDI market is the best example at the moment).

So it’s almost granted that every small company on the market will try to support at least these two platforms to survive, giving the Microsoft one a growing higher priority.

Is not clear instead how the ecosystem will interact with Citrix: supporting a third platform is a very expensive effort and in some ways the company is acting much more like a value added partner to Microsoft rather than like an independent virtualization vendor.
In this scenario small 3rd party firms may see Citrix as a competitor rather than as a preferred partner.

How much space is left for the other virtualization vendors?

Not too much. Parallels, Virtual Iron, Novell, Red Hat, Sun and Oracle will have to find a niche where their platforms make sense or will have to compete with Microsoft for the SMB.

Parallels is already doing that, strictly focusing on its OS virtualization technology which is the choice of preference all the time customers need a higher consolidation ratio at a lower price point. But the company is also releasing a server virtualization product: it has to be seen how Parallels Server will be accepted outside the Apple market.

Oracle is doing that as well: with a totally closed policy the company locks-in its customers obliging them to use its own hypervisor.

Novell is protected by the Microsoft interoperability agreement at the moment, so its implementation of Xen can be safely considered the Hyper-V of Linux shops.

Sun may be focusing on its own customers, so Virtual Iron and Red Hat seems the companies that most of all have to find a safe place as soon as possible. 
For both of them, using Xen as the virtualization engine, the young KVM platform seems a good opportunity to stay in the Linux community without having to handle with Novell and its powerful ally.
Red Hat already started to move in that direction.