Comatose servers -- also known as "zombie servers" -- occupy space and usurp power and cooling in a data center....
In today's data centers, 30% of servers fall into this classification, according to The Uptime Institute.
Zombie servers aren't really dead; they generally just use expensive resources without accomplishing anything useful. Often the only way to determine if it's worthwhile to keep running a server is to turn it off. If you find that it still has a purpose and supports workloads, start the process of moving its limited function to a virtual machine.
Why zombie servers are problematic
Consolidation and virtualization have reduced space concerns in data centers. Many enterprises have partially loaded racks simply because power and heat densities have risen to where full cabinet loads can't be supported. Some operators even consider old servers that fill up rack space, or useless racks that occupy floor area, as worthwhile because it gives the illusion of a full room and could keep upper management from reclaiming space. But these non-productive machines still use power, both directly to keep them running and indirectly in the form of cooling energy. And because many of them probably lack energy-saving features, the waste can be significant even at idle.
But it's not just the wasted energy -- although that should be a major concern in every facility today -- it's also the waste of valuable cooling resources that is an issue. Consolidated servers and virtualized platforms tend to concentrate loads in single racks or rack rows, resulting in high heat densities. Cooling these areas can be a challenge, and wasting cooling capacity on things that shouldn't use it can mean unnecessary capital expenditure on additional air conditioning. To make matters worse, the addition of cooling capacity without a careful study of its effects is often counterproductive. For example, if you add air conditioners in the wrong places, it can actually reduce cooling effectiveness even as it increases capacity.
Beat zombie servers with monitoring
Use a data center infrastructure management system (DCIM) to monitor multiple parameters, such as CPU utilization and power draw, and identify underused resources. If a DCIM isn't in your budget, you can create an automated tool to identify zombie servers -- as well as servers that are likely to become zombie servers -- and alert management staff.
Other steps organizations can take to reduce the risks of zombie servers include the using a configuration management database to track server utilization or appointing a dedicated staff member to continually measure and benchmark server utilization rates.
Consolidation in the data center takes years
Eliminate zombie servers to increase IT efficiency
Explore new data center efficiency metrics