Virtual servers are not free
Independent research in 2009 suggested that 15% of servers in data centers were not doing any useful work, and were costing $25 billion per year. Many organisations have found that this was an underestimate of the size of the emerging problem of virtual server “sprawl”, which results in unnecessary costs and wasted resources.
Organisations should pay more attention to retiring redundant servers in their cost-saving strategy
In October 2009 a global study by Kelton Research found that 15% of servers in data centers were not doing any useful work. They were not being detected by basic activity-monitoring tools because they were still performing housekeeping tasks such as updating their antivirus definitions. Extrapolating the study results globally indicated that they were costing their owners $25 billion per year in software licenses, power, cooling, and administration. Some IT managers were attacking the problem in a less than systematic fashion, using techniques such as turning off servers for a period to see if anybody complained! Virtualisation of servers doesn’t solve the problem without a strategy to prevent “virtual machine sprawl”
As we move into an increasingly virtualised data center environment, it is essential that we do not simply repeat the problem by generating and then forgetting masses of virtual machines. This problem is known as “VM sprawl”. Virtual machines are not readily visible and are often thought of as ephemeral. This makes VM sprawl difficult to control. However, virtual machines require software licenses, tie up storage, pose potential security threats, and generally waste resources, including energy. Redundant virtual machines obscure the view of the utilisation of their host processors and frustrate the efficient allocation of tasks to hosts.
VM sprawl has already become a problem in data center management. It is so easy to create virtual servers that users find it easier to create new virtual machines than to seek out an existing available asset, and it is also easy to forget to terminate a machine when its task is complete. Investing in tools that help monitor and manage these issues pays dividends.
Commentary courtesy of Ovum