In many data centers the motto is to leave servers where they are until the owner or manager of it requests a change. This is a common practice of supporting processes and applications with dedicated, unvirtualized servers that may or may not be performing work.
Then, add to this that most IT groups and data center managers are separate from the financials, meaning that the monthly utility bills are received and paid without much oversight. The economic pain of paying a higher bill than you need to doesn’t register since the impact is not direct.
But what if you knew that some of your servers were only utilized 12-20% and up to 30% of all of your servers were actually powered-up and sitting idle? Along with being a huge waste of energy and the associated cost these servers are also consuming cooling and space in your data center. Perhaps it may be time to do a bit of IT housecleaning.
However it’s not all that obvious or easy. Getting the approval to go through and remove or virtualize costs time and money and some may want to know what the return on that investment will be. Gaging against a years of operation, the cost savings will likely win; but also be prepared with other planned retrofits, changes and upgrades. Do the work and show how much the savings might be in kWh and $ with conservative estimates of how much time it may take your IT team.
A good place to start is with a basic inspection of the equipment. It may be surprising, but finding plugged in servers and storage that are unconnected does happen rather frequently. Correcting these simple mistakes make easy victories. Schedule surveys of the other IT assets to look for less obvious problems. These can be overcome by more crucial events, but they should be scheduled to keep them from dwindling out. There are services from vendors that can help, which can seem to shift the burden of risking a server outage but ultimately the IT group will still bear the responsibility.
Having an up-to-date inventory can help identify when servers were installed and which group may be responsible for them. Asking those groups to do an audit themselves, if they can, can be a great way for a group that hasn’t made changes in years to reassess what they actually need. Also groups change frequently and they may not know what they have; a reminder may be just what they need.
Despite utilization analysis, it can be just as much of a problem if you load up servers to 100%. This would introduce bottlenecks, latency, or other performance problems that limit effectiveness. That may be like a balancing act, but often there are servers that can ramp up to help with loads as needed and shouldn’t be considered unused. This also goes into knowing your applications and how they perform on a given class of server.
The latest issue also seems to be virtual machine ‘sprawl’, where more instances, images and virtual machines have come online because they are easy to deploy. But the impact can hurt, as they tie up disk space and energy. Archiving and consolidation can be key to helping with the sprawl.