Calling for capacity

In the first two instalments of this four-part series of Industry Insights on how to make the most of a virtual environment, I looked at the twin problems of VM sprawl and resource misallocation. By properly managing both the number of virtual machines in the environment and the resources allocated to each one, high and consistent performance can be achieved while still keeping to tight budgets.

In this third Industry Insight, I turn to the problem of capacity planning. A company's environment may be ticking over nicely for today's requirements, but what about six months down the line?

There's a tendency in the industry to dismiss the importance of capacity planning - it's not rocket science, is the usual line. But that line, if it was ever true, is not true anymore. In the days of virtualisation, when companies are always already over-committing their resources of memory, storage and processing power, capacity planning is rocket science.

Deciding what physical resources are needed to support the virtual environment is a complex business. And to make matters worse, while nowadays it only takes a few minutes to commission a new virtual machine, it still takes months to procure new hardware. If these needs are underestimated and capacity runs out, the company will lose the agility it was aiming for; if it gets overcautious and brings in too much capacity, the budget is shot.

How can a company accurately predict what its needs will be six months into the future - especially when virtualisation itself tends to increase demand from resource-hungry development teams?

The ability to draw on public cloud-based resources in a pinch can be handy. If a company is aware that demand is going to be subject to temporary peaks - at the financial year end, perhaps, or during a holiday sales season - it makes a lot of sense to use temporary capacity to fill the need. But, if the company prefers to keep it all in-house, or must cope with a genuine long-term trend of increasing demand, there is still a need to plan for, procure and commission new hardware on a regular basis.

If there was only one server and a small handful of VMs to keep track of, this task might be manageable without specialised tools. But, in an even moderately complex environment, a management tool that intelligently analyses trends and patterns in usage, and makes useful recommendations, becomes indispensable. A simple report that tells the company how much time it has before running out of memory, given current trends of usage peaks and troughs, or how many new VMs can be created with the current resources and usage, can make all the difference to a complex capacity planning exercise.

The latest tools on the market not only provide these smart capacity planning recommendations, they also enable what-if scenario planning. What if 10 new machines were added based on the current most-used virtual machine? What if the new machines were based on the profile of an existing machine in the accounting department, or a development team that tends to create unexpected surges in demand?

Capacity planning is rocket science.

This kind of scenario planning can also help to manage backup and business continuity requirements. What happens if one host is lost? Or all of them? Does the company have the headroom to cope with outages and failovers as well as support normal growth?

It's also useful to have a tool that maps virtual environments to the structure of the business. To most people, it will mean nothing to know that Cluster 1 is down - it's much more useful to know the HR machine is down. This used to be no-brainer information, but in a virtualised environment, everyone is sharing resources - half of HR's machines might in fact be sitting in a data centre on a different continent. Being able to name, categorise and group virtual machines - whether the categories are based on department, operating system, location or anything else that makes sense in the company - yields valuable information.

The combination of a business view and scenario planning is vital for companies that want the ability to plan capacity per department or per project, or to charge resources out to business units. A large new project can throw all long-range capacity planning calculations out of whack, and require the company to move up a purchase date - and decide whose budget to charge it to.

As with everything in virtualisation, getting capacity planning requirement right takes a combination of specialised tools and in-depth knowledge of the peculiarities of the company's particular environment. But once it is right, and the company has contained the problems of VM sprawl and incorrect resource allocation, it should have a virtualised environment that truly delivers on all the initial promises.

It's only at this point that most companies will begin to even consider moving their most important applications into virtual environments.

In the fourth and final Industry Insight, I will consider how to complete the final steps of stalled virtualisation programmes.