Jon Collins, originally published on The Register

No, no, not the v word…! There are few topics that have garnered as much interest recently as virtualisation. And that’s not just coming from us – the level of feedback we get on this is head and shoulders above many other areas.

But let’s be realistic. As respondents have informed us, the majority of workloads that have been virtualised to date are what we might class as ’easy targets’. This could be for one of a number of reasons – the workloads themselves are self-contained or low risk, or there is an immediate benefit to be had from virtualising it. And so on.

This is all far from the ‘vision’ of virtualisation, however. The potential of virtualisation goes way beyond the “take an application, migrate it to a virtual machine, move it onto another server” principle that we are seeing so much of today. We are told that server virtualisation can give us such things as:

More straightforward management of available physical capacity

Better control over the (virtual) machines themselves

Simpler operational management of the virtual server pool

Reduced risk due to more flexible approaches to business continuity

Ability to dynamically scale resources to meet fluctuating demand

Benefits such as these shimmer like a mirage against the very real, rather more clunky nature of today’s rack-mounted and big-box data centres. We know from previous workshops that, as virtualisation starts moving out of the pilot phase and into more mainstream use, you are starting to hit certain hurdles. Not least of these are bottlenecks with other parts of the IT infrastructure (can server virtualisation really excel without a compatible storage architecture, for example?) and issues around manageability.

All of which begs the question – what exactly will the virtualised server environment look like? We’ve seen some flagship examples from hardware vendors such as Intel and Dell, and while such organisations have to demonstrate their capabilities, we have also learned that the benefits did not always come quickly or smoothly. If wholesale adoption of virtualisation is not going to be an option, we can imagine several scenarios.

The first, of course, is that virtualisation remains in what is – to all intents and purposes, a niche. It could end up a very big niche – and indeed from a vendor perspective a very lucrative one – but nonetheless one in which virtualised machines sit in their own area, running their own workloads and applications.

The second scenario is that virtualisation will be a slow and steady burn, evolving into the mainstream rather than being any big bang. We’ve seen this happen for a number of technologies which needed so many other pieces to be ready before they could be fully adopted – business intelligence or iSCSI to name but two. Other pieces in the path of server virtualisation encompass management tools at one end, and appropriate hardware and protocols at the other – for example 10 Gigabit Ethernet and all that can ride on it.

Otherwise, we might see an avalanche effect, in the same way that we saw with mobile phones, email or the Internet. These broad areas were slow starters, but very quickly picked up when they went from a nice-to-have to an expectation. It’s not hard to imagine the scenario where it is just expected that virtualisation capabilities will exist, and if/when this happens it could be coupled with a certain sense of urgency – “Yes, I need to run this workload over on this server – and I need it now!”

These scenarios are not exclusive to larger organisations. For smaller companies perhaps the need is greater to bake virtualisation into the environment, such that the more standardised applications and packages can be run on a smaller, easier to manage footprint. For companies of all sizes however, whichever way the virtualisation journey takes us, it will require us to take into account what is already there. For some, ’legacy’ IT may be a ball and chain, for others it just works, and there is no reason to disrupt it. Taking existing IT into account requires virtualisation to be inserted very carefully as a layer – the technical equivalent of reversing the magician’s trick with the tablecloth.



Leave a Reply