Brian Madden makes some really strong points about the state of VDI today. To be honest (in my humble opinion) he is spot on when he says that VDI is a “use case” solution. He’s correct that even those customers that are 100% VDI today are companies where all their users meet the right “use case”.
He may even be right that come June 2010 perhaps VDI technology will advance so much that it can expand its “use cases” circle to include just about all desktops at all companies. But what if VDI doesn’t do that by June 2010? In fact what if VDI doesn’t do that ever? What would that mean for Citrix XenDesktop? And what would that mean for the competition’s VDI solutions?
To answer that, it needs to be pointed out that Brian frequently misses the boat when he talks about XenDesktop. In fact he recently assessed that XenDesktop is just an “old school server-based computing remote desktop delivery product”. To even suggest that, one has to seriously sell the XenDesktop product short of what it really is. And I find that surprising coming from Brian, as he has been one of the biggest fans of the Ardence technology, which is a significant component of XenDesktop.
I can only assume that Brian’s assessment of XenDesktop is based upon the use of XenDesktop as nothing more than a VDI solution. The problem with that basis is that XenDesktop is not just a VDI solution. It does far more than just VDI.
And this is perhaps the biggest problem that XenDesktop faces today. It is just so frequently compared to other VDI solutions, so much so that it’s now perceived to be “just a VDI solution”. It has pretty much become a pre-conceived notion, even within the ranks of the highly respected industry bloggers such as Brian Madden.
The problem with comparing XenDesktop to other VDI solutions is that it’s like comparing Guitar Hero to a Playstation 3. That’s comparing apples to oranges. In fact, that’s more like comparing apples to apple trees. Guitar Hero is a game. Playstation 3 is a gaming system, of which Guitar Hero is one of the games it can play. Likewise VDI is a tool, whereas XenDesktop is a Desktop Delivery system, of which VDI is one of the tools it can do.
So back to the questions in the second paragraph, what would it mean for Citrix XenDesktop if VDI use cases never expanded to include just about all desktops at all companies? And what would that mean for the competition’s VDI solutions?
If VDI use cases never expanded, then the competition’s VDI solutions would always remain “niche” solutions. However, because XenDesktop is far more than just a VDI tool, its technology is not contained within the VDI “niche”. To use the Guitar Hero/Playstation 3 analogy, if after this Christmas people stop buying Guitar Hero, that doesn’t mean people would stop buying Playstation 3. The gaming system isn’t contained within the “niche” of one of the games it can play.
The problem with XenDesktop today is that even though the technology isn’t contained within the VDI niche, its market perception is pigeon holed as being just another VDI product within the VDI niche. However, perception will eventually catch up with reality. Eventually the market will begin to understand everything XenDesktop can do. And if we are predicting what things will be like come June 2010, we can safely assume that the preconceived notion that XenDesktop is just another VDI solution will certainly be gone by then. If use cases of VDI haven’t expanded by then, then the market is still huge for XenDesktop, a product that does far more than just VDI. If VDI use cases do expand, then XenDesktop will still be right in the mix of that niche, as it does VDI as well.
OK, so what are these “other things” that XenDesktop can do, such that it’s a “Desktop Delivery” system as opposed to just another “VDI” product? Well, honestly I could refer to the many blogs Brian Madden has posted singing the praises of the Ardence technology, of which Citrix bought and renamed Provisioning Server.
In order to explain Provisioning Server technology, let me say that personally I’m not a fan of the product name “Provisioning Server”. Far too many times I have heard people say they don’t need it because they rarely provision operating systems. Such a task is something they do once to a new piece of hardware, and hopefully they never have to “provision” an operating system to that piece of hardware again. Well, “one and done” is not how Provisioning Server functions, but given the name of the product I can see why people assume it does function like that.
I much prefer to call the technology “Disk Image Virtualization”. Essentially what I can do with the technology is take a traditional desktop computer (or server) and virtualize the entire disk I/O that occurs on the machine. While the operating system is running, it’s using all the machine’s RAM, the machine’s CPU(s), the machine’s video card, the machine’s sound card, etc, etc… The only thing the machine isn’t using is the machine’s local hard disk. In fact, you could remove the disk if you want to, because the entire local disk I/O is being redirected to Provisioning Server.
So why not call it “Disk I/O Redirection”? Well, that would be a good name if that’s was all that it does. But Provisioning Server does one more important thing. It shares one disk image amongst many of machines. I could boot thousands of machines from one shared disk image. Calling it “Disk I/O Redirection” doesn’t address that it can share one disk image to many machines. So when I’m explaining what Provisioning Server does, calling it “Disk Image Virtualization” ends up making a lot more sense to the customer than the name “Provisioning Server” does.
Now again, Brian Madden loves this technology, despite what it’s called at any given time. The benefits of using it are huge. Instead of managing thousands of instances of Windows, I manage just one instance. Brian loved it when it was called Ardence. He loved it when it was renamed Provisioning Server. And I’m sure he’ll still love it even if he hates the name I call it, “Disk Image Virtualization”.
So I have to assume that if Brian assesses XenDesktop to be just an “old school server-based computing remote desktop delivery product”, he must not be aware that when a customer buys a XenDesktop license they are allowed to boot their local desktops directly from a shared disk image and avoid a remote connection to a VDI hosted desktop all together.
This flexibility of the XenDesktop license provides the XenDesktop customer tools for additional use cases, and more importantly, additional money saving implementations, which a customer of the competition’s VDI solution just flat out cannot do. Essentially, buying the competitions VDI solution for nearly the same price provides the customer with far, FAR less capabilities, and could end up costing the customer far, FAR more overall to implement.
Let’s take for example the case of the customer who decided to buy 150 band new laptops with OEM versions of Windows. I’m choosing the laptop scenario because I am frequently finding that due to the low cost of laptops today with OEM Windows installed, many companies are buying these laptops even for use cases where a desktop has been traditionally used. Their justification is that when they price out a desktop, or a thin client, or a laptop, the laptop is only costing a few dollars more than even some thin clients, so the cost to standardize everyone in the company on a laptop is pretty small.
If you’re pushing a VDI solution on the customer, the customer’s decision to buy laptops hurts your chances of selling VDI, because much of the savings in implementing a “cost saving” VDI deployment is accomplished by keeping the customer’s old desktop hardware on the user desks (i.e. save money by extending the life of existing hardware).
If the customer is committed to replacing the old hardware with something new, then the person pushing a “cost saving” VDI deployment would have hoped that the customer had opted for thin clients instead of laptops, so that the power savings of the thin clients could be included as part of the ROI on the VDI proposal.
Despite the customer’s excitement over deciding to buy 150 laptops, if you’re pushing a “cost saving” VDI deployment, then the customer’s decision is disappointing news to you. The new user hardware costs and power consumption of laptops will make your VDI proposal a tough sell (especially when the Windows license costs get added in… more on this later). You’d almost be inclined to convince the customer that buying 150 laptops wasn’t a good idea. Good luck with that.
But if you’re selling a Desktop Delivery system like Citrix XenDesktop, the customer’s decision to buy laptops doesn’t hurt your chances of selling XenDesktop at all, because XenDesktop has the tools to work with whatever the customer has on their user’s desks. Old, new, and/or power hungry, it doesn’t matter.
Keeping with the point of what “doesn’t matter”, if you’re pushing VDI on this customer, it doesn’t really matter to you at all how the customer intends to use those laptops, because no matter how they use them the customer is still going to need 150 virtual machines in the datacenter, and users are always going to connect via a remote protocol. But if you’re selling a Desktop Delivery system to the customer, how the customer intends to use the laptops actually matters (GASP!!! Imaging that… Use case matters!).
For this scenario let’s assume the customer says approximately 100 of the laptops will never leave the user’s desk, but 50 of the laptops could be taken on the road or home for the employees to remotely connect in. You also find out that they expect that the maximum number of remotely connected users at any given time would be 20.
If you’re pushing VDI as the solution, you would require a virtual infrastructure in the data center to host 150 virtual machines. But the person selling Desktop Delivery sees that the customer only needs the virtual infrastructure to host 20 virtual machines. This is a huge infrastructure savings, and given the customer just spent a lot of money on 150 laptops you’re going to need this savings in your proposal if you want the customer to buy a new way of managing their desktops.
Windows licensing can also be a huge factor. If you’re pushing VDI, the customer is going to have to take all 150 of those OEM Windows licenses and upgrade them to Vista Enterprise, add an annual SA fee, and then buy an annual VECD license. But the person selling Desktop Delivery sees that the customer only needs 50 of those OEM licenses upgraded to Vista Enterprise with SA and VECD. That’s one third of the additional Microsoft costs as VDI is only used when the user meets the VDI use case. See, use case matters!!!
Managing the operating systems is also significantly different. If you’re pushing VDI, you’re not giving the customer any tool to manage the 150 operating system instances that are on the laptops’ hard drives. All the tools of VDI only manage the operating systems on the virtual machines. Even if VDI finally develops an offline mode, the VDI tools still will only manage a virtual machine on the client, and not the operating system installed on the client’s hard drive. By pushing VDI, you’ve essentially just doubled the number of operating systems instances that this customer now needs to manage and you’ve only given them tools to manage half of them. Gee, thanks!
But the person selling Desktop Delivery tells the customer to have the 100 remaining OEM licenses migrated to the customer’s Microsoft Volume Agreement (which is much cheaper than upgrading to Vista Enterprise with SA and VECD), so that all 150 of the laptops meet the Microsoft licensing requirements to be booted from one shared Provisioning Server image (Vista Enterprise, SA and VECD are NOT necessary to use Provisioning Server… the operating system license just needs to be a Volume Agreement to use a shared image). Instead of having twice as many instances of operating systems to manage, the customer now has just one instance. The 150 laptops and the 20 Virtual Machines all boot from one shared disk.
And really the most significant benefit here is the user experience. When booting from Provisioning Server the user’s desktop experience isn’t occurring over a remote protocol. As much as I could sing the praises of ICA over RDP in a VDI solution, why even use a remote protocol when one isn’t necessary at all? When you push VDI, even when the user is in the office they’re using a remote protocol. Not necessary!
With Provisioning Server you’ve virtualized the entire disk I/O to occur in the data center where it’s secure, but do you really need all the CPU processing, graphics rendering, and multimedia processing occurring in the datacenter too? Those laptops have good ram, good processors, good video cards, good sound cards, etc… The user will be happier with all that processing occurring locally on their laptop. And IT will be happy that all of the user’s disk I/O is occurring in the datacenter on the Provisioning Server. Use a remote protocol when the user meets the use case. Don’t force it on them all the time.
So what about the 50 users who will take the laptop out of the office with them? What will they boot to? Well, first let me point out that they won’t be booting to the corporate image outside of the office. This is a good thing. In fact, that’s a great thing! It’s much more secure. In order to boot to the corporate image, a laptop needs to boot from the corporate network. If you’re pushing VDI those laptops need to have a corporate managed image on their hard drives because they connect to the corporate network every day. So when they leave the office with the employee, or they are stolen, they will boot with the corporate image anywhere that they are booted. Not good from a security standpoint.
Those 50 laptops in Desktop Delivery when not connected to the corporate network will boot with the image of Windows that the laptop shipped with. With the Citrix Access Gateway, and the XenDesktop Web Interface, the SSL encryption client and the Desktop Receiver client get installed automatically when the user remotely connects for the first time to the office. The employee can then securely use one of the 20 VDI instances within XenDesktop, which by the way boot from the exact same shared disk that their laptop boots from when they are on the corporate network. That means they get the exact same desktop outside the office that they have inside the office, but again the corporate image never physically leaves the datacenter.
In fact, if the user wants some offline applications, Citrix XenApp 5.0 is a great way to stream those offline applications over an http connection to them, and isolate those applications on an unmanaged operating system. And that’s essentially what that operating system on the laptop’s hard drive is. It’s unmanaged. You don’t need to manage it. It will never have a full connection to the corporate network. The only connection it will ever have is an encrypted ICA traffic connection.
If the user corrupts that local operating system on the laptop, just have them use the manufacturer’s restore CD and then point back to the Access Gateway and Web Interface to get the clients reinstalled. And if a restore CD is too hard for them, you’ve got 100 spare hard drives with the laptop’s original OEM Windows image on it that you pulled out of the other laptops that will never leave the user’s desks (assuming you don’t opt to sell them on eBay). You could have your remote users carry around a spare hard drive with them if you wanted to. In fact, you’ve got enough hard drives to have each remote user carry around two spare hard drives.
And ultimately, even if your company policy says you do need to manage the operating systems on those 50 laptops, that’s only 50 instances, plus 1 shared image that you need to manage, as opposed to 150 instances on the laptops and another 150 in the VDI infrastructure that the person pushing VDI creates.
Now, given the scenario of the 150 laptops, does XenDesktop sound like an “old school server-based computing remote desktop delivery product”? Sorry Brian, but you’ve missed the boat in your assessment of XenDesktop. I expect things will be drastically different come June 2010, but not for the same reasons as Brian. I believe by then the preconceived notions that XenDesktop is just another VDI solution will finally be gone.