Data centre managers need to get their house in order or face an energy crisis. Dramatic though this may seem, the solution is simple: basic housekeeping and strategic use of UPS (uninterruptible power supplies) says Robin Koffler, general manager, Riello UPS

Between April 2005 and March 2006, the total number of custom Apply er interruptions in the UK was around 21 million; total number of customer minutes lost 1,966 million.

Data centre association AFCOM predicts that over the next five years, power failures and limits on power availability will halt data centre operations at more than 90 percent of all companies.

According to Sun Microsystems, between eight and ten percent of servers in data centres have no identifiable function. The company announced in August 2007 it had cut the number of racks of its own servers from 95 to just five.

Ensuring power protection in modern data centres requires an upgrade and rethink in terms of infrastructure. Likely as not it will have the wrong type of power wiring installed under the floor and overhead and in insufficient number. The power distribution units may also be insufficient as indeed the quantity of breaker positions.

Separation of responsibility for power consumption between IT and facilities management has been a barrier to improvement. IT managers are interested in cramming in more server power. Energy consumption is rarely considered when purchasing new equipment. However, FMs pay the utility bill and thus have a much better grasp of the implications of high power consumption. They know about power loads and distribution of airflows but not necessarily IT equipment. They also have a greater comprehension of the wiring and electrical infrastructure in a building. The incentive for reducing power consumption should be split between the two.

To meet the challenges of next-generation, high-density data centres, UPS suppliers are following the lead set by computer manufacturers and placing greater emphasis on energy efficiency. Today’s units should have minimal impact on their environment and upstream power sources by offering:
• low input current distortion - less than 3%
• maximum input power factor - 0.9 or above
• power walk-in function to achieve progressive rectifier start-up
• delayed start-up on mains power supply return

There is a growing argument for pre-engineered even pre-manufactured UPS solutions that are modular, standardised, expandable and easy to install. The cost of advanced power management can be significantly lowered in this way. Challenges, with regard to UPS, are based around managing lifecycle costs, adaptability, scalability, manageability, availability and serviceability.

There has been a tendency to focus on current power continuity requirements only, resulting in over sizing, under utilisation and a ‘rip-out-and-replace’ scenario further down the line.

Such systems often required months of planning and bespoke engineering to make then fit the business, which is unacceptable in the data centre where planning for ten years of requirement and projections for power density are the norm.

Sizing a UPS system can be complex but in simple terms ranges include models rated in terms of kVA or MVA for very large installations. This is a measure of the amps their inverter assemblies are rated to supply at a set output voltage, typically 230Vac for single-phase and 400Vac for a three-phase system. The resulting value is a measure of the Apparent Power drawn by the load.

The core issue of sizing both the loads and UPS needed to power them (no matter which topology is chosen) is a crucial one and needs careful planning. Underestimating may result in future power disruption when forced to increase capacity and over estimation leads to higher initial installation and on-going maintenance costs.

Modular UPS systems, designed to grow with the requirement, will lower lifecycle costs and enhance serviceability. It also helps to minimise the number of circuit-breakers and wiring between the UPS and loads to negate potential points-of-failure.

Within a large critical installation, such as a high-density data centre, two or more distribution boards may be provided: one for the UPS and one for the generator that will support essential but non-critical loads in the event of a power supply failure. With dual corded equipment becoming the norm, N+1 electrical distribution systems have become the standard for data centre design. A true N+1 electrical design provides two separate and independent sources of power (A and B) to the load. With no interconnections or interdependencies, there are no single sources of failure that could impact both supplies. This provides incredibly high levels of reliability and is far easier to manage.

The power infrastructure that provides raw mains energy has changed little in the forty years it has taken for computing technology to replace manual operations. Like older parents know, the wrong side of 40 is a challenging place to be when trying to meet the voracious demands of a younger, more progressive entity. Data centre managers need to be ready with power continuity plans in place for when the creaking frame can’t deliver.