Time to get real about PC virtualisation
PC virtualisation allows multiple operating systems to be run on a single physical PC simultaneously -- effectively separating software from the underlying hardware.
If it lives up to its potential, the impact of this technology could be huge, according to Gartner. Any PC can be made to look the same to enterprise applications, irrespective of who made it and which OS is loaded as the host system, the analyst claims. The ramifications for Microsoft and the rest of PC industry are obvious.
ZDNet UK spoke to Brian Gammage, vice president and research leader at Gartner, about how virtualisation could revolutionise the PC industry, removing all scope for product differentiation and forcing vendors to compete on service and price.
Who is actually using this kind of virtual PC technology today?
It's been around for a while and it's in use in niche and technical applications. What is really changing is that it's now being deployed. The key player in terms of adoption are in the software industry, where the scenario is that you can run a machine that takes out hardware dependencies, and you can run separate windows to write code and test it.
We also see it in use by multiple-system users, people who would have multiple types of systems, such as traders in financial services. One early adopter in the USA is the National Security Administration (NSA) -- they have consolidated equipment that they definitely don't want overlapping.
I have also begun to see these technologies used by organisations looking to outsource. VMWare has a case study about one of the biggest insurers who use it to separate out the presentation layer in Windows and serve that to people outside the company with whom they have an outsourcing contract.
Microsoft has a key role to play here. What impact will virtualisation have on its software licensing model?
Microsoft bought Connectix because it wanted the technology. As software gets more complex migration costs go up, and in a saturated market anything that delays or inhibits replacement is a market limiter. Microsoft's big interest is virtualisation allows you to move to the new. You have to balance that against the licensing paradigm.
Legacy application support means moving from one OS to a new one, but if a few critical apps don't work on the new OS, that's a problem. A virtual legacy machine allows you to use those old apps and you don’t have to hold off the whole migration. And also, you can deposit your image of a PC anywhere you like. Longhorn is going to be the biggest migration since Windows 95 and it's going to be painful, so it's right to move now to deal with that. Market pressure is there. If you look at some of the Longhorn technical security features, they certainly give the impression that some support for virtualisation is built in.
How about Intel's role in this?
Both Microsoft and Intel have a vested interest in easing their user's migration path. Intel has announced hardware support to help isolate virtual machines and keep them secure. They will also have to allow privilege level support that will help you enforce installation and closure around a virtual machine.
What's the big pay-off for enterprise customers?
The big attraction is that deployment best practice is something that few organisations rarely achieve. Ideally you would lock users down, so you can intervene quickly. The measure of your exposure to a risk is the time from announcement to patch, and the more standardised your desktop the quicker you can intervene. We may actually see legal cases where companies who are unable to submit reports or deliver service because of infection and haven't followed best practice will get sued. And if people face litigation they will be under much more pressure to achieve deployment best practice.
What does this mean for the future of software licensing?
I think in the long term it drives a horse and cart through the current licensing paradigm. When Microsoft announced changes in product-use rights in October 2003, close to the date they launched virtual PC, they made it possible to install two versions of Windows on one machine. It's the first erosion of the licensing paradigm, and it takes out dual OS cost.
Assuming you play a double licensing fee, you could save 6 percent on TCO with a management host, and you get the management benefits of a standardised PC. Eradicating the cost of the additional licences becomes important. Microsoft is the big market decider; when they move everyone will follow. It's another nail in the current per instance licensing paradigm for desktop software. Other challenges are coming in terms of multi-core CPUs. We expect the approach to be extended over the next three years.
The current business model looks like it can't be sustained in the long run. Given the prospect of paying for two software licences most people will abstain. That's the closed loop.
How real is this today technically?
Functionally today, virtualisation can do everything we say. There are problems at the moment. The virtual machines look like apps which can confuse users. There are security issues, and virtual machines are kept as files in the host file systems, so you could delete the virtual machine. We think by 2005 this will be an attractive technology for bringing in those rogue IT users that IT departments are keen to get under control. It will take longer for them to roll out to everybody. It will be used to help people working from home, as a guest image of the work PC on the PC at home. Being applicable to everybody, we're thinking 2008 and 2009, we see mainstream migration to Longhorn, and we do believe that Microsoft will include a copy of virtualisation software with all volume licensing agreements.
You've also said it will help in outsourcing relationships?
It allows you to define the limits and boundaries of what the company who is working for you can see technically. It will be an important part of those IT support relationships - is the problem on the guest or the host system? It will assist the definition of service level agreements.