The pressure to go “green” in general, and in storage specifically, is an opportunity to assess and modernize IT infrastructures. The power savings from storage, while not overwhelming, are meaningful and can be used to catalyze organizational change.
By Dave Vellante
ccording to IDC, average data–center energy costs are growing at 20% per year and are expected to double by 2012. Discussions with Wikibon.org users confirm this expectation, as shown in the figure, right. Compounding this dynamic is the fact that power density is going up for all IT equipment, at a rate of 20% to 30% per year. This has the following domino effects:
- Much more power needs to be supplied to each square foot of a data center;
- More power is required to cool hotter equipment; and
- More heat extraction equipment needs to be supplied to each square foot of a data center.
The maximum heat density that can be air–cooled in a data center is approximately 10,000 watts/sq ft, yet most data centers were designed years ago for power densities of less than 1,000W/sq ft. The bottom line is that IT is approaching the limits of power distribution in many data centers, which is forcing organizations to begin exploring new cooling techniques such as water–cooled racks. Data centers are now constrained by power and cooling limits—not floor space, as was the case only a few years ago.
So where are the culprits in the data center, and how severe is the storage energy consumption problem? After all, storage is the fastest growing component of IT infrastructure for most companies. Is storage therefore the main problem? The answer is “no.” The figure on the right quantifies the culprits of power consumption within the data center based on interviews with users in the Wikibon.org community. The following findings are relevant:
- IT equipment directly accounts for more than one–half of energy consumption in the typical data center;
- Storage accounts for 15% of the total, while servers, at 30%, are the most consumptive;
- The current maximum loading for storage equipment is less than 1,000W/sq ft (compared to 1µ blade equipment, which is already at 10,000W/sq ft); and
- Projected storage loading is less than 2,000W/sq ft by 2012. (Migration from 3.5–inch to 2.5–inch drives is included in this projection.)
But storage is only one part of the problem in the data center. Within the storage stack, the main culprit is spinning disk drives, accounting for an overwhelming 80% of the power consumption of storage arrays on average, as shown in the figure on p. 35. (This figure will be lower for controller–intensive Tier–1 arrays.)
What can be done?
According to Wikibon member William Souder, director of network operations and chief security officer, Berry College, “There is no silver bullet when it comes to energy efficiency. Customers must evaluate a variety of alternative techniques to reduce power consumption and choose the ones that best fit their business requirements.” In general, with storage consumption growing at 50% per year for many companies, while budgets are flat, the best approach is to find techniques that help consume less storage and/or use storage more efficiently. Another approach is to get rid of equipment that is power hungry and to modernize the infrastructure.
Here’s a list of storage–related practices and approaches that can be used to reduce energy consumption, with associated trade–offs:
Virtualization and thin provisioning of server and storage infrastructures can provide immediate benefits, but they can bring complexities in management, performance, backup, and possibly a need to re–architect major processes in the data center.
Tiered storage can take advantage of high–capacity SATA drives and match data with device characteristics, but it’s not always easy to mandate that businesses will not be allowed to use Tier–1 storage unless they meet specific criteria (a.k.a. the “you’re not putting my data on Tier–2” phenomenon).
Data de–duplication, as applied to backup or online archiving, can reduce capacity requirements by, typically, 10x to 15x for backup or 2x to 3x for online storage. Often, however, disk–based backup using data de–duplication is designed to replace tapes, which are the greenest of technologies (notwithstanding the impact of transportation).
Massive array of idle disks (MAID) and spindown techniques that idle drives for inactive data are other solutions, but part of the problem is that organizations are increasingly wide–striping data across multiple disks for performance and resiliency. This means all devices are always active and can’t be idled.
NAND and flash memory are making inroads into conventional disk drives, and if you can cost–justify the technology it will save on power relative to spinning media.
In addition, there are several other approaches such as shutting down and consolidating smaller data centers (e.g., those less than 20,000 sq ft) but make sure you have top–down management support or the strategy will not stick. Selective sourcing (e.g., remote backup) is also an effective approach to lowering the power budget for small and medium–sized companies that are not leery of managed services. Other infrastructure considerations include DC power and water cooling; how–ever, these approaches are expensive to retrofit on existing infrastructure and are best applied in “greenfield” situations.
Finally, companies should take advantage of energy rebates and credits where available. To date, PG&E of Northern California is the only power company widely known to be writing checks to customers for installing more–efficient equipment.
However, according to Mark Bramfitt, program manager of the PG&E rebate program, “Not many storage vendors are taking advantage of the initiative.”
This does not bode well for the storage vendors and users. There’s something inherently wrong with the picture when vendors are all talking a good green game, but when it comes to something as fundamental as getting customers rebate checks, many have decided it’s not worth the administrative effort. This must change as all storage suppliers with a green story should be evaluated by their participation in such programs.
Think green, save green
IT organizations should use green initiatives as a lightning rod to catalyze actions around infrastructure modernization, which can lead to superb ROI (e.g., server and storage virtualization). Link green initiatives to corporate initiatives and market them internally.
Establish cross–functional teams and partnerships within IT and the business with representation from infrastructure groups within IT and the energy experts in facilities management. Bridge the two perspectives and begin designing power and cooling considerations into virtually all decisions, understanding the trade–offs and benefits.
Vendors can be useful in this process and bring proven track records and tools such as power calculators, design expertise, and best practices.
Finally, treat green initiatives like a normal project with clear measurements, solutions, and timelines for reduction in power consumption. Communicate results as you would in any project, set expectations, and strive for continuous improvement. Perhaps most importantly, make green “cool.”
David Vellante is a co–founder and contributor to The Wikibon Project (www.wikibon.org), a research and advisory community of practitioners dedicated to the open sharing of business technology knowledge.