Cutting the power draw while ensuring data reliability and availability are the prime considerations in modern day retrofits and new construction. With those twin goals in mind, it's no wonder that more enterprises are taking another look at battery power for critical infrastructure.
In recent weeks, top tier organizations like AT&T and Fedex have unveiled sizable battery backup systems at key data centers, with the intention of maintaining full-load operations for as long as it takes to restore either utility or on-site power . At the same time, new tools and techniques like variable-frequency disk drives help stretch that power window as far as possible by dynamically tailoring energy consumption to actual data requirements.
Backup is one thing, but is the industry ready for all-DC power? Emerson Network Power's Mark Murrill thinks so. He notes that DC power requires far fewer conversions than AC on the way to actual IT equipment, so both the complexity and energy footprint of the data center's electrical infrastructure are greatly reduced, perhaps by as much as 20 percent. He also notes that row-based DC power units can be hot-swapped on the fly, virtually eliminating the mean-time-to-repair (MTTR) factor that contributes to AC failure.
The emergence of new low-power server, storage and network architectures is also making battery power more compelling. A key development is the new ARM technology hitting the channel. Marvell, for example, just came out with the 1.6 GHz Armada XP quad-core device, which uses barely a quarter of the power of equivalent Xeon architectures without even factoring in SoC extras like Ethernet controllers. The device is also Energy Efficient Ethernet compliant and employs DDR3 memory that provides a 20 percent efficiency gain over DDR2.
No matter what architecture or power infrastructure is in place, however, beware of rosy claims when it comes to UPS performance, according to APC by Schneider's Shri Karve. Impossibly high efficiency claims usually mask problems in source switching as the system struggles to maintain supply to the server infrastructure. That, in turn, increases the risk of failure at crucial moments -- offering protection that is barely superior to a low-grade consumer unit.
It's somewhat ironic that just as IT infrastructure is transitioning from an in-house to utility model, power supplies are poised to come back off the wire in favor of local generation.
It hasn't happened yet, but a number of signs are pointing in that direction.