It can be tempting to think of virtualisation as a magic potion that will automatically allow IT departments to do more with less – but, says Warren Olivier, regional manager at Veeam Software, really making it work requires some changes in attitude, approach and tools.
“An organisation that tries to manage virtualised IT infrastructure in the same way it managed the old hardware-based servers, with the same tools and habits, is going to be frustrated and disappointed,” says Olivier. “You need a different mindset, a new set of skills and new tools.”
The first big mistake newly virtualised organisations make, says Olivier, is to think they can carry on using their existing tools for backup and monitoring.
“Agent-based tools live inside a machine and report on what they see – they’re inside the fishbowl. But the whole point of virtualisation is that individual virtual machines (VMs) have no idea they’re virtual – each one thinks it has access to the full resources of a physical machine. Any information you gain from inside that fishbowl is going to be inaccurate, especially when it comes to resource utilisation.”
Some applications are now being designed to be VM-aware in a process called “paravirtualisation”, says Olivier – but this doesn’t change the need to invest in monitoring and backup tools that are specifically designed for virtual environments.
“Getting the right tools in place is non-negotiable,” says Olivier. “Without them, organisations won’t only lose out on some of the benefits of virtualising, they’ll actually find that things become more complex and difficult to manage – a step backward.”
The second key mindshift, says Olivier, is to understand that “a virtual environment is no place for equality. Some machines are more important than others, and you need to establish a hierarchy and treat them accordingly.”
IT managers trip up, says Olivier, when they fall into the habit of creating all their new VMs off a single common template. “If you do it this way, a lot of the time you’ll end up either under-allocating resources – in which case you’ll have performance problems – or over-allocating them, which is wasteful.”
To get resource allocation to each VM right, Olivier says it’s important to track patterns and trends in usage, and match allocation to that. “The good news is that virtualisation allows you to reallocate resources at any time, so once you have the right information you can improve performance very quickly.”
Managers should beware, he says, to use representative sample periods: “Anything less than a year can be dangerous, especially when it comes to machines that are involved in year-end financial processes.”
With the right tools, says Olivier, “all this really is easier. Without them, you’re in trouble. No virtualisation project is complete until the most useful monitoring and backup tools have been researched, budgeted for, purchased and installed.”