Like many large companies, Intel's computing needs are growing fast -- it expects server usage to grow 15 percent to 23 percent per year and storage needs to grow 40 percent per year.

But over the past two years, through virtualization and grid computing, Intel has achieved an 11 percent increase in server utilization, which will allow it to save $77 million in 2008 capital purchases. To reap the remaining cost-saving initiatives, Intel is:

Consolidating data centers -- Intel has 130 data centers in 31 countries, but many of them are small; nearly 30 percent occupy less than 1,000 square feet. The company plans to turn half its data centers into offices or labs and run large hub data centers with better cooling. Intel already reduced its number of data centers by 13 percent in 2007.

Consolidating servers -- As you might expect, Intel is replacing all of its servers with multicore Xeon-powered servers on a regular four-year refresh cycle (in some cases even less than four years), rather than let old servers linger on. It has already reduced its number of servers by 1,800 while increasing the overall number of cores. Virtualization, Intel hopes, will allow it to consolidate the workloads of 15 to 20 older servers onto each new Xeon-based server. Intel also plans to to implement a global grid that will let semiconductor design teams use idle servers at remote data centers and thus increase server utilization from 66 percent to 80 percent or better.

Sophisticated storage -- Intel is deploying tiered storage and information lifecycle management systems that should reduce storage costs by migrating less-important data to cheaper media. It's already using self-service provisioning that lets users request their own storage; when their requests are in line with IT policy, they're assigned the requested disk space.

Using green computing -- The company plans to use a data center heat recovery system that will be part of Intel's first "green" building. Heat produced by servers in the data center will be used to warm offices and provide hot water. "We expect that this approach will save energy, accumulate points toward green certification, and deliver annual cost savings estimated at hundreds of thousands of dollars," the report says.

For those of you who want to make your own data centers more efficient, Coy Stine, thermal management engineer for Degree Controls, offers the following tips:

1. Separate hot and cold airflows. Implement hot aisles and cold aisles, install covers for cable cutouts and blanking panels in unused rack space

2. Eliminate coldspots. Utilize cold air more efficiently, allowing the production of less cooling.

3. Conduct a Computational Fluid Dynamics Analysis (using software to map the entire data center, seeing where all the air flow is) and implement changes based on the results. A lot of companies perform such analyses, find excessive cooling in one area of the data center, and don't do anything about it.

Greg MacSweeney
In a Matter of Months, CIOs’ Agendas Turned Upside Down
Early in the summer of 2007, Wall Street was counting its good fortune. In early summer 20...

Larry Tabb
Clearing and Settlement Top-of-Mind for Front-Office Execs
In addition to the excitement in the U.S., the real focus is occurring in Europe, where th...