In this article, I want to discuss some of the issues surrounding the assessment of various virtual desktop vendors from a technical and licensing perspective. Following discussions with my colleagues in the field, it has come to my attention that there may be a possible mole -- in the shape of a totally new virtualisation vendor -- gaining footing inside the virtual data centre.
Six months from now, it is likely that VMware View and Citrix XenDesktop will have feature parity. Right now, the spanner in the works for many is that VMware's much-vaunted PCoIP protocol is incompatible with their security server, along with the fact that Citrix historically maintains an edge over VMware as a client-friendly, firewall-friendly connectivity suite.
When this feature disparity is narrowed, however, a more significant disparity might become more apparent. At the moment, both VMware and Citrix make it cost-effective to have a closed model where you deploy just their solution, both for the hypervisor and the virtual desktop "broker."
Of course, XenDesktop is compatible with ESX, and that might be of interest to existing VMware shops with plenty of spare capacity on existing clusters. But what if there isn't spare capacity? What if you are buying new hardware for a new virtual desktop infrastructure (VDI) project and it is too big to be slotted into an existing VMware configuration as an afterthought? What if functional and structure differences make it more desirable to have a set of physical boxes running VMware and server-based virtual machines (VMs), and another set of physical boxes running Xen and XenDesktop? Why might pick you such a configuration?
It could be that you are uncomfortable with an amorphous block of physical servers running a blend of workloads comprising production, test/dev and virtual desktop VMs. Indeed, despite the fact that creating "silos" of servers for different tasks historically sprang from limitation of the physical world, the silo approach has benefits from a workload management, performance and permissions perspective. But setting all these soft benefits to one side, the thorny issue of cost might make you consider another vendor.
There are plenty of people, most of all Microsoft, who will regularly tell you that "VMware costs way too much." That is not my argument here. I have a more subtle point than just price, which is this: Virtual desktops represent a unique workload, making them significantly different from the server consolidation workloads that virtualisation and VMware introduced in the last decade. Virtual desktops are cost-sensitive in a way that perhaps server VMs have been sheltered from.
This stems from the fact that desktop compute has always been cheaper than server compute. Back in 2003 and 2004, it was very easy to create a value proposition surrounding consolidation. It's much harder to do this with virtual desktops when the licensing models don't reflect this difference. We have always been trying to do more with less, but with virtual desktop we are trying to do even more with even less. Currently, the licensing models don't allow for a blended solution (XenDesktop+ESX) or technically, they do not support it (View+Xen). That's going to force customers to do a more like-for-like comparison of the vendors for their virtual desktop projects than perhaps they did for their server consolidation projects, which were started when there was only one serious virtualisation vendor in town -- VMware.
The "just enough hypervisor"
Many of my colleagues are beginning to tout the idea of a "just enough hypervisor" specifically angled around building virtual desktop solutions. For many, the way vendors have created their SKUs make for a blended solution more expensive and over-featured for the job in hand. The type of hypervisor needed for virtual desktops is light, slim and not overblown with unnecessary features. For many of my colleagues, Xen and XenDesktop currently fit the bill. In fact, they see decided advantages in introducing a different virtualisation vendor into the environment -- by keeping server VMs and desktop VMs in separate environments, they limit too much dependence on a single vendor.
Additionally, they see this as an ideal way of playing one vendor off against another, as they currently do with hardware purchases. Put simply, if you only run one single vendor's virtualisation platform, it's pretty hard to convince your contacts with the virtualisation vendor that you are serious about assessing the competition.
There will be some that say this deviates from the principle of one-throat-to-choke, and that going down the multi-vendor route will be inherently more complicated and expensive. But given that we already work in a multi-vendor environment anyway, along with the fact that you might be able to save significant sums from a licensing perspective, the money saved could be ploughed into better hardware and training your staff to be more than just a VMware one-trick pony.
This could have consequences for VMware. Effectively, some might say they have unwittingly opened a window that Citrix could use gain a foothold within the data centre. Once customers are using and supporting one of its competitor's products, they will be better informed about its limits and capabilities, and much more prepared to undertake a move. It's a small chink of light in the data centre, which could allow Citrix and Microsoft a virtualisation foothold they so desperately need. From there, it's a much smaller leap to consider running not just virtual desktops on Citrix XenServer, but server based VMs as well.
The desktop, not the hypervisor
Perhaps Citrix's recent move to make XenServer open-source and free to download now makes more sense -- it's an attempt to boost their market share with the hope that folks will be willing to give their XenDesktop solution a try. At the end of the day, all an end-user cares about is their desktop, not the hypervisor that powers it.
Psychologically, this approach might be harder for VMware to adopt, given its financial and engineering investment in ESX. Many of its peers identify this as a situation where VMware wants to have its cake and eat it, too. In one breath, they are told that future is about management, not the hypervisor but management, and yet VMware still attaches a premium to their ESX product. VMware has made a great play of the superiority of their hypervisor compared to their competitors. But with the much improved hardware we now have available, these enhancements, particularly for virtual desktop loads, are less a unique selling point than they once were.
Personally, I think the reality is that the cost and performance differences between VMware View and Citrix XenDesktop are pretty small. The standard list prices breakdown as follows:
- VMware Enterprise View: $150 (£97) per user
- Citrix Enterprise XenDesktop: $225 (£145) per user
- VMware Premier View: $250(£162) per user
- Citrix Platinum XenDesktop: $350 (£227) per user
Back in the real world, it's worth mentioning that the bigger the deployment, the bigger the discounts can be negotiated. Most large corporate accounts can save significantly on licensing, and the biggest discounts come from Citrix.
As with all licensing, however, the costs are not comparable because the two vendor's features never match pound-for-pound or dollar-for-dollar. They frequently include different features that defy attempts to place a cost value on the feature. For the moment, buying VDI solutions remains very much like buying a new mobile phone -- a dizzying array of tariffs and add-ons constructed in such a way as to prevent cross vendor comparisons.
My own personal opinion on this matter is that much depends on the organisations existing commitment to VMware and/or Citrix. It is true that the VMware View product is clearly less sophisticated and less refined, but if you're an existing VMware customer with spare capacity on your ESX clusters, the View product offers good value in conjunction with an easy setup.
If, on the other hand, you're planning a large VDI project and need a feature-rich environment with an experienced vendor in client delivery, then XenDesktop is worth evaluating. Indeed, strategically it might be well worth the investment in maintaining a relationship with Citrix. This is especially true if the organisation is already a Citrix MetaFrame/ Presentation/ XenApp customer, as then XenDesktop easily integrates into an existing server-based computing environment.