Too hot to handle? Power and cooling is always a hot potato
Jan 23rd, 2007 by Rick Barnard
Even though it is the middle of winter (at least in North America) and oil prices are at a 12-month low, people are still talking about energy. The new Congress announced it will focus on energy independence and sources of alternative energy – with goals of reducing the cost and environmental impact of producing energy. And this isn’t the first time either. Last year (July 2006), when energy prices were at their highest, a bill was created to study and promote the use of energy efficient computer servers in the United States. Nothing ever came of that, but expensive energy is here to stay and it’s costing companies billions.
Here are some stats to back that up:
- There are more than nine million servers running 24 hours a day, seven days a week
- 38 percent of the U.S.’s power supply is being absorbed by data centers (Source: Univ.NH study)
- Businesses will need an additional 12 million square feet of data center space by 2009 (Source: IDC)
If the last statistic proves to be true, this means that businesses are going to spend anywhere between $1 to $3 billion per year in utility costs to power only the additional systems that they need. And, building the datacenters to put them in may increase from $5 billion to $60 billion.
From all of this – it is clear to me that one root cause is the increased processing density in the datacenter. The end result is datacenters are filled with hundreds to thousands of servers and all of the supporting infrastructure including network switches, storage networks, cables, and cooling systems. Each of these components requires power, cooling, and real estate which drives up operational costs for organizations.
So, how can one reduce the processing requirements that businesses need to run their applications? There are several ways with the goal of increasing your application density:
- Increase the utilization of the processing resources
- Eliminate the number of processing resources in the first place (e.g. passive for HA/DR)
- Reduce the amount of supporting infrastructure (e.g. network switches, disk drives, etc.)
Many companies are leveraging the benefits of what we call “data center virtualization” to reduce their energy costs. Virtualization at the data center level can reduce the number of servers required for an application by 50% (while maintaining service levels to the customer) and eliminate 80% of the components that support the application.
With the example above (just the incremental growth in datacenter requirements), this would translate into billions of dollars in annual utility savings and billions more avoided by not building new data centers.
Just wait until the dog days of summer.