Now that Citrix is focusing on end-to-end application delivery, people have been speculating / asking whether Citrix needs to get into the hypervisor market. (And when people ask “Should Citrix get into the hypervisor market,” that’s really code for “Should Citrix buy a hypervisor?”) In fact when we were at Citrix iForum in Edinburgh, Scotland this past June, we had multiple conversations with people (both Citrix employees and others) about whether Citrix should own a hypervisor.
(A hypervisor, for those who don’t know, is a software virtualization engine. It puts the “VM” in “VMware.” It’s what allows multiple virtual hardware machines to simultaneously run on a single physical system. It’s what Microsoft is building into Windows Server 2008 to allow multiple server VMs to run at once.)
There’s a rumor that Citrix is going to buy virtualization vendor XenSource for a very large amount of money—perhaps as high as $500M. In this article, we’ll look at why Citrix would want to do this, what it would mean for them, and what it might mean for their existing relationships with Microsoft and VMware.
Why would Citrix want to own a hypervisor?
Everyone knows that Citrix has grown from a pure server-based computing company into an application delivery company. Virtual desktop delivery (whether you call it VDI, xDI, or VCC) is a big part of that.
Citrix released Citrix Desktop Server (CDS) v1.0 this past April. Even though it was a brand-new product in terms of SKU, CDS shares a lot of code with Presentation Server. At iForum Edinburgh this past June, Citrix publicly announced some goals for future versions of CDS; specifically, the ability to connect to Windows desktop VMs directly via the ICA protocol and the ability to manage pools of Desktop VMs as if they were single-user Presentation Servers.
This VDI/xDI/VCC thing is hot. Even if you don’t believe in it today, the concept of delivering a desktop as a service is going to continue to grow. (In fact, many people are now realizing that some form of this might soon replace all desktops in a corporation—-not just the “special case” scenarios that are popular today.)
The problem is that if you want to do this right now, you need products from several different vendors. Some vendors (such as Virtual Iron and Provision Networks) have partnered to offer single-SKU end-to-end desktop delivery solutions that combine the virtualization, desktop brokering, and server-based computing management into a single platform. The problem with these non-Citrix solutions is that remote access to the desktop VM is provided via Microsoft’s RDP protocol, and the reality is that in today’s world, RDP does not perform nearly as well as ICA in WAN environments. The SpeedScreen technologies, the compression, and the graphical performance of Citrix Presentation Server in just flat out blow away RDP across WAN links. (In reality, the Provision / Virtual Iron solution is a “local network” solution only.)
Citrix sees the value of delivering desktops. They have / will have this great ICA / Presentation Server-based product for doing so called “Desktop Server.” They have a way to manage images with Ardence. They can track performance with EdgeSight. They can stream applications into the desktop with the application streaming capabilities of Presentation Server. Citrix has a complete solution except for one thing: a hypervisor.
The official line from Citrix about this is that they’re “hypervisor agnostic.” But this is more of a marketing spin phrase of convenience as opposed to a specific intentional move from Citrix.
But should Citrix be hypervisor agnostic? Remember that Citrix calls Desktop Server a “DDI” solution, which they define as a solution can be used to deliver three types of desktops: multi-user (Terminal Server + Presentation Server), blades, and VMs. This is all well and good, but the reality is that as server power increases, people are not using blades for desktop delivery (except in extremely specific circumstances). Citrix already owns the multi-user shared desktop with Presentation Server. So why not also own the whole solution stack in the VM desktop space? Citrix is currently hypervisor agnostic because they don’t own a hypervisor.
Now imagine if Citrix did own a hypervisor. They could sell a single product that would truly deliver desktops to users no matter where they were. The customer would only have to provide their own Windows images to make this all happen—Citrix could handle the rest. (And again, Citrix components like Ardence could greatly simplify this process.) Thinking like this, it’s easy to see why Citrix would want to own a hypervisor too.
Fine, so Citrix wants a hypervisor. Now what?
If Citrix wants a hypervisor, they could build their own from scratch, they could OEM license someone else’s, or they could buy someone. Building their own is not really realistic in today’s world. Licensing one might lead to long-term risk. (Just look at the risk they took when they had to renew the Microsoft Terminal Server source code agreement.) That leaves “buy one” as the only viable option. And as a company doing over $1.3B a year in sales with a market cap pushing $7B, Citrix can certainly afford it.
So that leads us to “which one?” The two that come to mind immediately are XenSource and Virtual Iron.
Citrix tried to make a strategic investment in Virtual Iron. They wanted to offer them a bunch of money, and in return they asked Virtual Iron to work exclusively with them in the desktop delivery space. Virtual Iron refused. Citrix went back to them with more money in an attempt to buy Virtual Iron outright. Again, they refused.
So Citrix moved on to XenSource. XenSource is an interesting company. There is an open source hypervisor on the market called “Xen.” XenSource is a commercial company that was formed to enhance Xen (kind of like how SuSe and Red Hat enhance the open source Linux for commercial gain). XenSource is particularly interesting because Xen and Microsoft have a fairly solid relationship. In fact “Mike Nell, product unit manager for Microsoft's virtualization technologies...pointed out that Microsoft Research in the United Kingdom contributed to the development of the Xen hypervisor technology, which was initially a project at the University Of Cambridge's Computer Laboratory.” (source: eweek)
Over the past few years Microsoft and Xen have collaborated on a few more occasions. Microsoft engineers helped to enhance the Xen hypervisor. Xen licensed the VHD virtual hard disk format from Microsoft not too long ago, and most recently Microsoft and Xen announced an agreement to enable Xen-enabled Linux guest VMs to run on Viridian servers with the full support of Microsoft tech support.
Ian Pratt--the founder of XenSource--is the inventor and project lead for Xen. So it’s conceivable that Citrix could acquire Xen and not piss off Microsoft. (In the short term at least.)
Long term, this could be tricky for Citrix. First of all, you gotta figure that it’s going to take Citrix a few years to really integrate that into CDS. But during this time, Microsoft will continue to develop their own hypervisor (“Viridian”). So when Citrix gets dialed in with XenSource, will that even still be relevant if Microsoft has a super hypervisor built-in? Will people still want/need a Citrix hypervisor? Would there be value in that? If so, what?
But more importantly, imagine how strategically important Viridian will be to Microsoft? Even though Microsoft and XenSource are friends now, what do you think will happen as soon as Viridian is released? All of these "partner" hypervisors will becoming "competing" hypervisors overnight. It will be "kill, or be killed." (After all, this is not the same thing as Citrix Presentation Server adding value to an obscure Windows Server feature. This will be about citrix replacing a key strategic Windows Server feature.)
With risks like this, why is Citrix even looking to buy a hypervisor? And why are they looking to spend so much money on a company that only has a few million dollars in sales?
First of all, Citrix probably still feels burned that they didn’t get to by Softricity.
Second, there is always value in owning the whole stack. The problem with Microsoft developing a hypervisor is that Citrix couldn’t control when / if certain features came out and how they would integrate with the other Citrix components.
A final complexity to this whole thing is around Citrix’s systems management story. Citrix has traditionally been very careful to avoid getting into the systems management space. But now that they have streaming application delivery and EdgeSight, if they add a hypervisor and some management tools, they might be treading into new territory there too.
An interesting side note to this whole conversation is VMware. Citrix has explicitly named them as a competitor in recent months, and if this deal goes through, it would absolutely square the two off as real competitors. (Throw a little Cisco investment into VMware and you have some real competition!)