Thursday, June 03, 2010

Can Virtualization Drive Savings? A Quick Estimation Method.

An associate asked recently if there was a quick way to estimate the potential value of virtualizing a datacenter. The following model should help organizations quickly estimate the potential cost savings, and illustrate why datacenter virtualization is currently the highest priority project for most organizations.


In a typical datacenter, the spending is allocated as follows:
  • Management and Support Labor = 38%
  • Hardware = 20%
  • Software = 30%
  • Facilities (power / cooling and space) = 12%


Virtualization's first order impact is to the hardware costs, helping to consolidate servers (10:1 is typical average today, although higher consolidation ratios are available), and reduce storage costs with SANs, reducing headroom and overprovisioning by 25%. Overall virtualization can reduce hardware costs (acquisition, support contracts) by an estimated 40% overall.
Facilities costs are reduced with consolidation of hardware, and modernization with more energy efficient servers, helping to reduce power, cooling and space costs by 45%-50%.
Net software costs are harder to reduce, as virtualization software / management tools usually require an incremental software investment. However, when virtualization occurs reconciliation of licensing is usually performed at the same time, and some physical server based licensing can be avoided, helping to offset the incremental virtualization software investment.

Management and support costs are also reduced via streamlined provisioning and reduced change costs, estimated to reduce datacenter management and support labor by 15% to 20% overall. Improving management tools and practices maturity while virtualizing can yield much higher improvements of 25% or more.
Tallying these costs overall yields an annual 20% moving from a non-virtualized to virtualized datacenter.
What can this mean per application? With $6 to $9K per application per year typical, annual savings of $1,200 to $1,800 can be expected.


No comments: