In a commentary on Forbes.com earlier this week titled: "Servers: Why Thrifty Isn't Nifty," Kenneth Brill, executive director of the Uptime Institute, made an alarming statement:"We are currently in the biggest data center construction boom in history. At the same time, this boom is dramatically weakening the future flexibility and financial performance of information technology. ...The number of servers in the U.S. has grown from 5 million in 2000, to 10 million in 2005, to a projected 15 million in 2010. More servers eat up more electricity and energy costs go up. To avoid future energy shortages caused by increasing IT demands, 10 more power plants need to be built to the tune of $2 billion to $6 billion each, and their cost is ultimately going to get passed on to IT through increased utility bills."

The Uptime Institute is in a position to know about data center efficiency since it monitors data center uptime and maintains the nation's largest database of how data centers physically fail. Uptime's May 2008 report, Revolutionizing Data Center Efficiency, identifies three major problems that wreak havoc on efficiency in most data centers: 1. Poor demand and capacity planning within and across functions (business, IT, facilities) 2. Significant failings in asset management (6% average server utilization, 56% facility utilization) 3. Boards, CEOs, and CFOs are not holding CIOs accountable for critical data center facilities CapEx and data center operational efficiency.

Uptime proposes a three-part solution to double IT energy efficiency by 2012 and to arrest the growth of green house gas emissions from data centers:

1. Mandate inclusion of true total cost of ownership (including data center facilities) in business case justification of new products and applications to throttle excess demand 2. Rapidly mature and integrate asset management capabilities to reach the same par as the security function 3. Formally move accountability for data center critical facilities expense and operations to the CIO and appoint internal "Energy Czars" with an operations and technology mandate to double IT energy efficiency by 2012

Although the Uptime report recommends virtualization technology to improve utilization and reduce the number of physical servers when planning data centers, it doesn't mention cloud computing, an issue I explored in a previous blog. Brill asserts that most organizations have no idea what their facility cost for a single server is:

"For an organization with 5,000 servers, the industry rule of thumb is that up to 30% are technologically obsolete. This means that up to 1,500 servers can just be unplugged with no negative impact on data-center production. The savings: $12 million to $23 million recovered in data-center facility capacity, $700,000 in annual electric savings, and 6,000 annual tons of reduced greenhouse gas emissions. ... If we did this on a broad national scale, do we really need to be building all the new data centers, or could we defer a large portion of this investment into the future? Our companies and economy would be far better off if that money went into new application development instead of bricks and mortar!"

Virtualization and cloud computing technology can help with these kinds of server obsolescence and utilization problems since virtual data centers and the ability to scale up or scale down cloud servers is no doubt on track to becoming a viable technological option, although few companies seem willing just yet to run their mission-critical apps on a data center in the clouds.

Massively Parallel Solutions?

Another energy efficient solution not explored in the report is massively parallel processing architectures. Data center servers used for general-purpose computing are dominated by Intel/AMD, or x86, architectures. This chip technology is at the heart of blade servers and rack servers, including those used in most data warehouse applications. There exists, however, a new generation of data warehouse appliances, such as the Netezza Performance Server (NPS) system, that uses an asymmetric massively parallel processing (AMPP) architecture specifically designed for high-performance data warehousing, which allows it to consume significantly less power and generate less heat than other solutions. Netezza's architecture uses intelligent storage nodes to process data as it streams off the disk, thus increasing performance. Each of these nodes uses an embedded PowerPC chip that consumes 4.5 watts of power (compared with 70 or more watts in x86-based systems), reducing the power typically required by the vast majority of data warehouse solutions.

Data center managers can learn a lot from Brill's Forbes commentary and the Uptime report. In addition to heeding Uptime's advice to take into account all relevant costs for data center servers, as well as turning off or replacing old and inefficient servers, I'm convinced that a strong case can be made for a few data center pilot projects willing to explore alternatives to the typical x86 architectures, especially if your company's goal is to eliminate or defer new data center construction.

For more information:

1--See my previous blog entry Reducing Data Center Energy Consumption.

2--In a special report, InformationWeek Analytics looks at how increasing requirements for processor cycles, memory, and storage as well as higher electricity demands are affecting data center costs--and offers some tips on how to design a modular data center that will future-proof your investment. The report can be downloaded here.