TECHNOLOGY FOR CHANGE

Reducing Costs Through Energy-Aware Routing

Salvatore Salamone | Date: 08-26-09 | Comments

Companies have used load balancing for years, and virtualization and cloud technology more recently, to shift computing workloads to optimize performance and make more efficient use of computing resources. Want to take this approach to the next level?

Companies have used load balancing for years, and virtualization and cloud technology more recently, to shift computing workloads to optimize performance and make more efficient use of computing resources. Want to take this approach to the next level? How about factoring in kilowatt-hour pricing fluctuations around the country or the world to optimize your computing�s electrical costs, too?

That's the idea behind an idea called energy-aware routing where workloads would automatically be shifted to data centers that offered the lowest price per kilowatt-hour at a particular time.

A new study by researchers from MIT, Carnegie Mellon University and Akamai found that large businesses using this technique potentially could reduce energy costs by as much as 30 to 40 percent. (A paper detailing the group�s work was presented at this month�s SIGCOMM 2009 conference in Barcelona, Spain.)

By studying electricity prices for 30 U.S. locations for three months (January 2009 to March 2009), the researchers found �a significant amount of day-to-day volatility, short-term spikes, seasonal trends and dependencies on fuel prices and consumer demand.�

Examining hour-to-hour price fluctuations in few locations, the researchers found that in New York and Chicago the minimum and maximum price during a single day can easily differ by a factor of two.

To take advantage of this pricing variability would require several elements to be in place. First, a company would need real-time information about electrical prices for all of their data centers. Second, an intelligent management system would need to correlate current workloads, energy usage and location-based electricity pricing and make decisions as to which sites offered the optimal pricing per workload. And third, there would have to be a mechanism in place that would allow workloads to be shifted seamlessly from location to location.

Looking at these requirements one by one, cloud computing and virtualization technologies as they are used today would help satisfy the third condition of being able to shift workloads. Real-time electricity pricing information is becoming more readily available as the utilities deploy their smart meter and smart grid technologies. The sticking point is the intelligent management system. This is something that still needs to be developed. But given the level of energy awareness and management already integrated into enterprise systems' management systems, the ability to optimize use based on workloads, energy usage and electricity�s pricing does not seem like science fiction.

That said, there are potential obstacles that might prevent or diminish the potential benefits of energy-aware routing.

Currently, many companies are locked into contracts with their utilities where they agree to buy a certain amount of electricity for their data center at a fixed rate. If a company with such a contract wanted to take advantage of energy-aware routing, it would have to change the terms to allow the variable usage this approach would require.

Additionally, energy cost-savings in shifting workloads might be neutralized by other costs. For example, a company might incur higher bandwidth costs when constantly moving compute jobs (and the associated data) around the country or the world.