Dale Vile, originally published on The Register
The benefits of virtualisation, particularly in relation to x86 server consolidation, are pretty well recognised. The significant reduction in hardware requirements and operational overheads that many have achieved with their initial deployments will in many cases have easily justified the cost of the virtualisation technology itself.
As organisations move beyond the low hanging fruit of obvious consolidation opportunities and begin to work virtualisation into the fabric of their infrastructure, however, both the cost of acquisition and the ongoing total cost of ownership come into sharper focus.
Fortunately, from a supplier perspective, the market has moved on from the early phase in which x86 server consolidation was just a one horse race, with VMware pretty much having the market to itself. Today, we have a couple of other significant players in the running, namely Citrix and Microsoft, and with everyone chasing the mainstream opportunity, we are already seeing the premium prices associated with the early market being challenged.
With hypervisor bundling and open source options also in the mix, there is then the question of whether at least basic virtualisation capability should be something we have to pay for at all.
Beyond the basics, however, we need to consider how virtualised environments are managed, particularly as deployments are scaled up. One of the big advantages of virtual machines is the speed and ease with which they can be created and provisioned, but this is a double-edged sword. Some have found, for example, that the problem of sprawl quickly raises its ugly head.
This not only starts to reintroduce some of the operational overhead that virtualisation was designed to get rid of in the first place, but can also make it difficult to control things from a software license perspective, potentially creating both cost and compliance issues.
Picking up on this last point, it is also clear that some platform software, middleware and application vendors are still struggling to deal effectively with some of the licensing and support issues that arise as we move from deployment on physical to virtual machines. The practice of tying software license fees to physical machine capacity, for example, creates obvious issues when consolidating workloads from several smaller boxes onto one larger one. Situations like this can easily lead to customers being penalised unfairly and/or running into yet more compliance exposures.
And coming back to the operational question, there is then the additional cost of having to recreate problems on a physical machine before support can be obtained from software vendors who refuse to acknowledge the legitimacy of running their product in a virtualised environment.
Perhaps the biggest operational challenge, however, is creating the most appropriate infrastructure for monitoring and managing virtual machines. In all but the smallest of installations, the requirement here is for cost-effective systems management solutions and there are two schools of thought emerging in this area.
The first is based on the premise that virtualisation brings with it a unique set of operational requirements that are best dealt with through a dedicated systems management approach that works in parallel to your existing management tools and processes. The second argues that a more holistic approach to systems management is more effective, so the requirement is for existing tools to be extended so they can be used to manage both physical and virtual assets together in a coherent manner.
Either way, there is a potential cost implication in terms of incremental investment required.
In addition to the above, when we consider the expense of training or hiring to acquire the necessary skill sets, the upgrades that are often necessary to the storage and networking infrastructure etc., it is clear that we need to think beyond how much we pay, or don’t pay, for basic enabling software such as the hypervisor when counting the full cost of virtualisation.
Against this background, we would be interested in your views and experiences with regard to the true cost of virtualisation, particularly as you scale up deployments.