POWER SUPPLY Improving uninterruptible power supply availability and energy efficiency
Given how seriously a power failure could impact a modern data center and its clients, it’s not surprising that most operators protect their facility’s critical equipment with an uninterruptible power supply (UPS). Modular topology allows UPSs to fulfill their vital role with maximum availability as well as good energy efficiency and scalability.
On 27 May 2017, British Airways suffered a global IT systems failure when a data center went down for around a quarter of an hour. The resulting chaos saw tens of thousands of passengers stranded, luggage dispersed, and aircraft grounded.
The problem was attributed to a ‘power supply issue’.
While not all data centers are the size of BA’s, scenarios like these motivate most to protect their critical IT load with uninterruptible power supplies (UPSs). Online UPSs assure the quality of mains power when present, and seamlessly switch over to battery back-up if the supply fails. The battery autonomy should be enough to allow a generator to start up and take over the supply, or, failing that, for the load to shut down safely.
It follows that the UPS itself must be as highly-available as possible; in today’s business and social climate, it should also be energy-efficient, and easily scalable to the current and future needs of its supported load. One way of helping UPS systems to meet all these objectives is to use modular UPS topology.
Modular vs. monolithic UPS topology
We’ll start by explaining what modular topology is, and then show how it delivers its benefits. Originally, UPSs were built with a monolithic design - and some still are. These supported, say, a 400 kW load with a single, floor-standing 400 kW unit. However, with advances in power semiconductor technology, and the advent of transformerless designs, it’s possible to build complete UPS units as small, rack-mounting modules. Accordingly, our 400 kW capacity could be fulfilled by a cabinet populated with four 100 kW modules. If the cabinet could accept up to six modules, then one or two more could be added at a later date for instant scalability.
Availability and redundancy
However, modular topology also contributes to availability in two ways, because improving availability depends on maximizing mean time between failures (MTBF) and/or minimizing mean time to repair (MTTR) according to the equation:
Availability = MTBF / MTBF+MTTR
One way of increasing MTBF is to use more reliable—and more expensive—components in the UPS’s construction. However, there is a finite limit to component reliability, irrespective of costs. Another, better, solution is to build in redundancy; this can yield a three- to six-fold improvement in MTBF for power protection devices.
For the monolithic system, redundancy can be achieved by using two 400 kW units to support a 400 kW load; if one unit fails, the other can continue fully supporting the load. However, this means that, during normal operation, 800 kW UPS capacity is being provided; at best, each UPS is only running at 50 percent capacity. If the load drops further through lower demand, the UPSs may be driven into an energy-inefficient operating region.
By contrast, an extra 100 kW module can be plugged into the modular cabinet for N+1 redundancy. Five modules, of total 500 kW capacity, are supporting the 400 kW load. Each module is 80 percent rather than 50 percent loaded, so far less likely to be driven down into an inefficient operating region. Yet, if any one module fails, the others can still fully support the load.
Accordingly, this approach allows redundancy to be set up, and availability to be improved, far more efficiently. Excess capacity, capital costs, cooling, and floor space requirements are all reduced.
Smart module switching enables Modular UPSs to run efficiently
One caveat is that as module numbers increase, MTBF can reduce due to the higher component count and potential for failure. However, this is more than offset by the modular system’s reduced MTTR. A faulty module can be withdrawn from the UPS cabinet, and a replacement plugged in, within about half an hour. This compares with six hours typically needed to repair a monolithic UPS in situ.
The impact this has on the availability equation means that modular UPSs with availabilities of 99.9999 percent, sometimes known as ‘six nines’, are now possible.
Modular UPSs can be kept running efficiently, even if they’re subjected to a load reduction. This is achieved by a technique called smart module switching, where the number of active modules is adjusted according to load requirements. Surplus modules are switched to standby, but remain ready to start again if the load increases. Efficiency improvements become especially significant if the load drops below 25 percent of UPS capacity.
Overall, modular UPS system advantages accrue from their flexibility and ability to support rapid, incremental changes.