I’ve been talking to some companies that have been pursuing projects to “get the PC off the desktop.” There are different ways to do this, which have their good point and bad points, but first, why would you want to?

I explained the fundamental technology motivation a few months ago in a posting entitled Why The Desktop Is Broken. The fact is that PC support costs an arm and a leg. Exactly how big an arm and how long a leg varies, but a reasonable rule of thumb is that the annual Total Cost of Ownership (TCO) of a PC can be a 5x multiple over its cost. With the right management software you can pull the cost down, but the evidence strongly suggests that if you get the PC off the desktop, you cut the TCO dramatically, by about 50 percent. So you could save a whole limb.

In another posting, Does Client Virtualization Make Sense?, I listed four ways of pushing the PC into the data center and leaving nothing more than a thin client on the desk. (Thin clients are essentially screens with keyboards and just enough local processing power to handle screen and keyboards activity.)

What became clear to me from talking to IT users out there is that two of the four techniques are very complementary:

1. The PC Blade: as implemented by IBM, HP and ClearCube. This is where a whole PC (except for disk store) resides on an unshared blade in the data center. There will be some local disk for booting Windows, but storage will be via a Storage Area Network (SAN) or Network Attached Storage (NAS)
   
2. Client Virtualization: as implemented using VMware’s hypervisor (but other hypervisors could be used) and VMware’s Virtual Desktop Infrastructure (VDI). Here virtual PCs (i.e. whole instances of Windows) are created on the server or server blade. So the user has what appears to be a complete PC on the server, but in reality it may only have a fraction of the resources of a full PC. If this is configured correctly then the user will never know, because the hypervisor will provide all the power the user needs. In this technique the PC’s disk store is provided by a SAN

The truth is that some PC users – the so-called power users – really do need a whole PC. Even if they only run one application like, for example, Photoshop, that requires (in brief bursts it has to be said) as much CPU as you can provide, and which is also a memory hog, they will either run that app very slowly or it will disrupt the activities of others. The outcome depends on how you configure the virtual PCs that share a blade or server.

Even when you have, say, 15 virtual PCs sharing a single blade all of which have dependably low resource needs, you can run into trouble if, for example, each PC launches an Anti-Virus update and check at the same time.

There are some other disadvantages to the VDI approach which limit its application to some degree. The primary one is cost. Putting 15 virtual PCs on a blade sounds like a big win in terms of hardware spend, but remember that PC hardware costs have never been the big deal – and both a PC blade and VDI need a thin client, connection software and management software. Also, currently VDI solutions actually need a SAN, whereas PC blade solutions can make do with NAS, which is less expensive. If you don’t happen to have a convenient and compatible SAN, which you can add capacity to, then VDI could turn out more expensive “per PC user” than PC blades. I’ve spoked to IT users who abandoned VDI implementations because of this, and moved to PC blades.

Alternatively, if you have just completed a server consolidation exercise and you have a mostly idle server farm to find work for, and the SAN is no problem, then VDI could prove highly economical – and you may evade the need for any PC blades for a year or two. However, remember that the hypervisor itself can chew up a good deal of resource and that you almost certainly have some power users who really do need power. If most of the PC population is going to disappear into the data center, you’ll almost certainly need PC blades.

Let me emphasize again that the big one-time win is in reducing PC support costs, and both solutions do that, because they get the PC off the desk and drop it into the data center. Now stir in the fact that PCs tend not to be used at all for a good part of any 24 hours and you can then think in terms of making those resources, whether dedicated PC blades or shared resources, available to be used elsewhere.

It’s easy to conceive of a symbiotic merger happening between these two technical approaches. All that”s required is a well-thought out partnership between VMware and a PC blade provider like HP, IBM or ClearCube. This could be the way that the IT industry is headed as regards solving the desktop problem.

This is a posting in the Virtualization Focus Series. Click here to see an index of such postings.