UPS systems exist solely to deliver clean, uninterrupted power to their critical load – but their ability to do so is entirely bounded by their resilience to failure of their own components or subsystems. In this post, we look at how today’s UPSs achieve the availability essential to data centre operations everywhere.
A UPS must operate with the highest possible availability, where availability is defined as the relationship between Mean Time Between Failures (MTBF) and Mean Time To Repair (MTTR), formally expressed as:
An ideal UPS system would have 100% availability, meaning it suffers zero downtime in any given period. 99.9999%, usually referred to as ‘six nines’ availability, represents a downtime per annum of 32 seconds, while ‘five nines’ equates to 5.5 minutes. By contrast, 99% would mean 87.5 hours’ downtime a year, so an apparently incremental improvement is actually of critical importance.
However, modern UPS systems with modular topology are achieving ‘six nines’ availability levels, and they’re doing so by improving both factors in the Availability equation – MTBF and MTTR. This is ultimately made possible by the transformerless UPS technology used in today’s static UPS designs.
Transformerless UPS designs are popular because they offer significant energy efficiency improvements over the earlier transformer-based designs – and they maintain this high efficiency across a wide load spectrum. Meanwhile, though, transformerless UPS systems bring another critical benefit; a significant reduction in both size and weight. This means that a UPS solution that was previously implemented as a large, floor-standing installation now comes as a manoeuvrable UPS module that can rapidly be plugged into or removed from a 19” racking frame.
Accordingly, a 60 kVA load (for example) could be supported by three 20 kVA UPS modules in a modular UPS frame. Alternatively, to achieve improved availability, the UPS system can easily be incremented to four modules, providing 80 kVA capacity in an N+1 redundancy mode. As a redundant system remains available even after an individual component failure, its availability is improved compared with a capacity-only UPS implementation.
There can be a downside, though. Achieving a redundant solution from a single multi-module rack is more attractive in terms of cost and floor space than two floor-standing UPS units in a 1+1 implementation. However, any system with multiple UPS modules will have lower reliability than a redundant UPS with just two units, because of the increased number of components inevitably used.
Yet a modular UPS system will still be the best solution if it has hot swap capability. This is because if a module fails in service, it can be removed and replaced without the need to work on the UPS system in situ, or even power it down. As a result, MTTR is reduced from the six hours typically needed for a legacy system to just half an hour for a module replacement. This, in turn, means that modular UPS systems with both redundancy and hot swappable capability can achieve the 99.9999% availability essential to data centre infrastructure today.
A UPS installation with high availability forms part of the bigger picture that will be on the mind of any ICT manager or owner; the availability of his entire data centre. This issue is addressed by the Uptime Institute, which has developed a Tier Classification System for data centre availability and specifies a set of attributes for each tier level. More information on this is available within KUP’s UPS Handbook, or on the Uptime Institute’s website.