Walking the Line Between Data Center Efficiency and Reliability
There’s a saying that you can never be too thin or too rich. In the data center, that could be converted to: You can never be too efficient or too reliable.
But while thin and rich do not necessarily work at cross purposes, the same can’t be said for efficiency or reliability. The more you improve the former, it seems, the more you diminish the latter, and vice versa.
Yet, the drive for greater efficiency remains at fever pitch throughout the enterprise industry, with nearly all of the activity focused on reducing energy and infrastructure costs while maintaining readiness to handle the cloud, Big Data, mobility and the myriad other forces affecting the data center.
The first thing many IT executives fail to understand about efficiency is that it doesn’t necessarily require vast sums or herculean efforts to achieve results, says Emerson Network Power’s Jack Pouchet. In most cases, outfitting just the core systems with low-power components and advanced power management tools will benefit the broader data environment. The key to making it happen is establishing a clear roadmap that will coordinate efficiency measures, both large and small, and ensure that efforts in one area are not diminished by those in another.
It’s also true that not all data environments are the same, so efficiency initiatives will vary as well. EBay, for example, recently devised a new metric to gauge the effect that changes in its IT infrastructure will have on its e-commerce activities. The Digital Service Efficiency model calculates the number of transactions taking place per kilowatt hour. The company has devised its own dashboard allowing managers to keep track of the metric, and has even offered it up to The Green Grid as an industry standard.
The U.S. Department of Energy, naturally, has a lot to say about data center efficiency, even going so far as to build its own high-performance computing center optimized for high efficiency. The National Renewable Energy Laboratory (NREL) is a showcase for advanced energy systems integration and renewable resource utilization to show that even petascale operations can cut energy usage without compromising productivity. Through techniques like warm-water cooling and high-voltage rack support, the agency is aiming for a power usage effectiveness (PUE) of 1.06 or better, which is about as close to 100 percent efficiency as is realistically possible.
At some point, however, efforts to improve power efficiency will have to focus on the power supply itself. This is where the efficiency/reliability conundrum becomes most acute because most data centers rely on over-powered UPS infrastructure in order to be ready for, well, anything. But as Eaton’s Janne Paanen points out, new modular UPS designs make it easier to scale power loads up and down. Couple these with the latest Data Center Infrastructure Management (DCIM) solutions, and enterprises of all sizes will have unprecedented ability to dynamically match power supply to data requirements.
Like beauty, efficiency is in the eye of the beholder. Every improvement to the data/kilowatt ratio will be met with polite applause (maybe) and then a request to see the plan to achieve the next level. This isn’t a bad thing, mind you, in that it fosters high demand for data infrastructure specialists.
But it does mean that data center efficiency will not likely be “solved” for some time to come – neither within the industry nor in the eyes of the public.