Growing Cynicism Around Going Green
Evidence is mounting of a growing cynicism regarding green initiatives within the IT infrastructure space. We may be reaching a point where vendor hype has hit a saturation point and beginning to meet with customer resistance. While there is a genuine concern about data center power consumption, particularly with regard to accommodating increasingly dense technology footprints, the larger concern for most, particularly in the current climate, is controlling costs.
One of the sources of frustration with vendor-driven green initiatives is that true efficiency improvement requires a holistic effort. The efficiency of one device is often impacted by associated elements in the supply chain, all of which may be dependent on external factors, such as positioning and layout for cooling within a data center.
Perhaps there's a lesson to be had from the diesel automobile, now poised to be reborn in the US market. Led by the German marques, these vehicles, long popular in Europe but excluded from the U.S. due to an inability to meet emission control standards, will begin reappearing in the 2009 model year. The new breed of diesels meets the most stringent pollution standards and offers better fuel efficiency than their gasoline counterparts while maintaining comparable performance. Interestingly, they accomplish this through a combination of technologies that each provide incremental improvements that when applied collectively produce a clean, efficient, high performance vehicle.
As with diesel cars, the key to success is to consider all of the components and examine the optimization potential. Currently, servers have been responsible for significantly more power consumption than storage, so logically this has been where the greater focus has been. As organizations improve server efficiency through a combination of virtualization initiatives, upgrades to more efficient technologies, and better management, they will turn to storage.
Challenged with the need to store and protect ever-increasing quantities of data, it is not surprising that power costs for storage has increased over the past five years at a faster rate than servers and network. (It should be noted that some of this increase may simply reflect a transfer of power usage as companies add storage arrays as they virtualize servers, reducing server power consumption by eliminating locally attached disks and adding them to arrays).
While still representing a relatively small (but growing) percentage of overall storage acquisition and operation costs, at some point improving storage power efficiency becomes significant. Today, there are two primary approaches to accomplishing this: better utilization and reducing the number of spindles. The former can be done through a combination of policy (e.g. monitoring, classification, and chargeback) and technology (thin provisioning, deduplication, etc.). The latter is accomplished through tiering and redistribution of as much data as possible to high-density storage that can be further enhanced via power-saving techniques like massive array of idle disks (MAID).
Unless and until some dramatic breakthough occurs, a strategy that blends the right combination of these techniques remains the key to storage power efficiency.