Published on Friday, Dec 18 2009 by
While IT managers seek to maximise computing power from limited floor space, they are also tasked with controlling energy consumption, emissions and costs. Alan Luscombe of Uninterruptible Power Supplies Ltd (UPSL) explains how - in addition to protecting critical data centre loads - advanced UPS technologies can help meet these challenges.
With IT power consumption up 400 percent per server rack since 2003, and little sign of a slow-down, data centre managers and designers have a tough task specifying UPS systems that will cater for future needs. Pressure to shrink carbon and physical footprints, plus rising electricity costs and economic constraints, further complicate the equation.
Nevertheless this is a pressing matter. Exponential growth in Internet applications, processing power and data storage is outstripping the level of protection existing UPSs were originally designed for, meaning that many legacy systems are incorrectly sized for today’s needs and insufficient for continued load expansion. Further, inefficient systems burn excess electricity and create needless heat emissions, compromising environmental objectives. There is also a real risk of failure in a climate where power continuity for mission critical loads is arguably more important than ever.
Modular solutions
The drive for data centre efficiency, flexibility and availability has been key to the development and uptake of modular UPS solutions. The scalability of modular architecture can deliver major reductions in electricity consumption and CO2 emissions, and help specifiers to make flexible plans for power and space requirements for both immediate and changing future needs.
Trying to cater for future power requirements with traditional stand-alone UPS systems can lead to over-specification, creating a wasteful gap between installed capacity and the size of the actual critical load, and making inefficient use of costly floor space. However, rack-mounted configurations can be right-sized by inserting or removing ‘hot-swappable’ modules, enabling power to be added as requirements grow without any footprint penalty. This hot-swap technology, along with significant reductions in repair time, can also achieve six nines availability (99.9999 percent) - highly desirable for data centres in pursuit of zero downtime.
Modular, transformerless UPSs, with Decentralised Parallel Architecture (DPA), provide a flexible, space-efficient and transferable system, versus static stand-alone installations that may never be used to capacity and would certainly be a challenge to relocate.
A smaller footprint
Energy and environment considerations are coupled with the high costs of real estate, particularly in city centre locations, and this has emerged as a major incentive for data centre managers to seek space savings for their IT systems and ancillary equipment. Demand for power can lead to plant being larger than the data centre it is supporting, so it goes without saying that any contribution to space saving is to be welcomed.
For example, a floor space reduction of 70 percent could be achieved by replacing a 10 year old 400kVA parallel redundant UPS system (running at 45 percent of its rated capacity) with a new DPA 200kVA parallel redundant UPS system.
A succession of space-consuming IT deployments, combined with property design and layout restrictions, can impose physical constraints on subsequent server installations and supporting infrastructure, especially in old or converted locations. Modular rack-mounted transformerless UPS systems offer a flexible and space-saving solution for such challenges while meeting exacting performance specifications, since they provide high power density (a class-leading 340kW/m2 for the latest DPA systems) and the smallest physical footprint on the market. Compared with legacy systems, such modular UPSs typically take up only a quarter of the floor space.
The cost of failure
While major investments have been made in expanding and upgrading IT systems, many are still reliant on single stand-alone UPSs with Centralised Parallel Architecture (CPA). While this sharing of common components may offer some cost benefit, the centralised configuration introduces a number of ‘single points of failure’ into the system, which adversely affect its availability. Decentralised systems are uniquely designed to remove single points of failure, achieving virtually zero downtime and the elimination of costly disruptions to mission critical operations. The additional capital expenditure on a DPA system will be recouped through availability gains and by providing enhanced protection against revenue losses caused by system failures.
Decentralised Parallel Architecture works by paralleling independent rack-format UPS modules. This means that each individual module contains all the necessary hardware and software required for full system operation. With all critical components duplicated and distributed between the independent modules, potential single points of failure are eradicated, giving guaranteed system uptime.
With a minimum of one module over and above that required by the ‘capacity’ system, the load is supported with UPS power if any one module shuts down, thereby providing full N+1 redundancy and significantly increasing system availability – an important factor at a time when power supply in the UK is becoming less dependable but more critical to business operations.
Cost savings
Modularity improves efficiency by working closer to the load capacity than traditional UPSs but without sacrificing the security of the system. The more a load approaches the capacity of any UPS, the more efficiently the UPS operates. A traditional stand-alone parallel redundant system is typically just 50 percent loaded while a modular solution typically achieves a 70 percent or higher loading. This reduces both energy and UPS cooling requirements.
Savings in annual running costs, emissions and floor space can be achieved, and more than compensate for the initial purchase premium of an advanced modular system. For example, approximately £150,000 could be saved over five years by replacing a ten year old 400kVA parallel redundant UPS system, running at half of its rated capacity, with a new DPA 200kVA parallel redundant UPS system. This would also reduce CO2 emissions by over 700 tonnes and cut floor space by 70 percent.
The scalability of modular systems also contributes major savings. Compare a single stand-alone non-redundant 100kVA UPS solution with a parallel redundant 3 x 50kVA UPS rack-mounted solution. While the latter may carry a price premium, the cost-benefit is quickly apparent. The modular configuration provides redundancy if one of the units fails, and spare ways can accommodate an increase in capacity in affordable, incremental steps without interruption to the critical load. The stand-alone system provides no redundancy, and the addition of a second parallel 100kVA unit to increase capacity would be more costly, take up twice the space, and would also incur downtime during installation.
When compared with conventional parallel protection systems, decentralised modules reduce electricity costs, heat loss and CO2 emissions, while delivering the industry’s smallest footprint and N+1 parallel redundancy.
The impact of blade servers
Data centres are dynamic computer environments and the increasing mix of old and new computer technologies is causing the overall power factor of servers to shift towards unity. Furthermore with the introduction of powerful blade servers the overall power factor may even become leading.
This creates a problem for legacy UPS installations. These mostly use pulse width modulated (PWM) technology which provides maximum power into lagging power factors.
As the load type changes from lagging to leading, these UPSs quickly approach their kW power limits or even go into overload. Replacing the existing legacy UPS with a higher output unit is expensive and usually disruptive, with changes in the power distribution and installation. Adding a second such UPS incurs even higher cost plus extra demand on premium floor space.
Instead, the best approach uses a modern transformerless UPS with adaptive inverter switching. This derates far less severely than a legacy double conversion system as the power factor moves to leading. For example a 300kVA transformerless UPS could deliver 232kW into a 0.9 leading power factor load whereas a legacy double-conversion type would supply just 182kW. In other words, the transformerless UPS has derated by just 3 percent with respect to its rated power at a power factor of 0.8 lagging, while the legacy system has derated by 24 percent.
This improved efficiency means that a lower power system can be specified. Improved operating efficiency of UPS systems reduces the burden on air-conditioning plant in computer centres, thus making a further contribution to energy efficiency. Transformerless design also imposes a reduced total harmonic current distortion on its input (THDi). Eliminating the transformer makes the UPS considerably smaller and lighter as well.
There is huge potential to manage and reduce the electricity consumption of data centres, and alleviate the burden on cooling systems, by continually matching the capacity of UPS systems to their respective critical loads. State-of-the-art modular, transformerless UPS systems are much more flexible than their traditional counterparts at matching load requirements and delivering optimum efficiency, enabling data centre managers to cost-effectively meet their power performance, space planning and environmental objectives without compromise.
Alan Luscombe is sales and marketing director at Uninterruptible Power Supplies Ltd (UPSL), Aldermaston, Berks.