The development of UPSs as critical data centre components
24Mar
posted by: Kenny Green, Uninterruptible Power Supplies
Incoming raw mains is fed to a rectifier/charger for conversion to a DC output for the inverter and the UPS battery charge
Kenny Green, Uninterruptible Power Supplies, considers how users’ expectations of UPSs have developed since they first appeared, and how today’s users can benefit from 30 years of UPS technology evolution.
When UPSs first appeared in the mid-seventies, they serviced computer installations that aided productivity but were not always critical to their users’ business survival. Today, with 24/7 real time, on line transaction processing, this is no longer true; a data centre failure would be catastrophic.
The UPS (uninterruptible power supply) plays a vital role in preventing such catastrophes. The UPS has changed radically over the last three decades, not only because of the technology used, but also because of the changes in their working environment and above all, how their role is perceived by operators.
Back in the seventies, organisations’ computing resource - when they had any, was concentrated into large, centralised and monolithic computers. With desktop PCs not yet available, distributed users had only remote terminals with no inbuilt intelligence. Centralising the computing resource – an early version of a data centre, made the provision of the UPS and other essential services easier.
Role change
However expectations related to the UPS’s role were more relaxed than they are today. This was partly because the role of the computers they protected was also different. At that time, they were novel, and seen as a welcome improvement to productivity in selected parts of an enterprise, rather than a vital part of the entire operation. A computer failure would cause delay and inconvenience, but it was unlikely to drive an enterprise permanently out of business, as it could do today.
Attitudes to energy efficiency were similarly relaxed, with few concerns particularly about energy costs; companies were not facing the legislative or political pressure to preserve energy and ‘go green’. In addition, computer installations and their UPS were not of the size and power that they are today.
During the eighties, the rise of the PC made distributed computing extremely popular. From the nineties onwards, IT discipline began to re-assert itself as networking, client-server architectures and then Internet communications became the norm. Companies felt compelled to establish an Internet presence and became increasingly dependent on digital technology to handle transaction processing in real time and on line. Larger organisations established their own centralised data processing and communications facilities, whilst smaller concerns entrusted their data handling to outside agencies. Either way, this resulted in large facilities containing banks of tens, hundreds or even thousands of servers - all contributing vitally to the existence of the organisations using them. The age of the modern data centre had arrived.
2G data centres
Users’ expectations of these ‘second generation’ data centres differ from those of their predecessors. Firstly, computing capability changed from a novel option to a vital resource. Secondly, energy efficiency has improved. With sharply rising energy prices, poor efficiency creates a financial burden. Additionally, companies can face pressure from government, shareholders, customers and employees if they cannot demonstrate a responsible Green strategy. Another point is that modern blade servers allow data centres to be far more flexible than earlier large monolithic computers in responding to rapidly-changing demands.
UPSs contribute critically to helping data centre operators overcome these issues. They prevent power failures or disturbances to the mains reaching vulnerable ICT equipment and causing irrecoverable equipment damage or loss of data. UPSs with the right topology protects with high efficiency, minimising both electrical losses and cooling costs. They can also offer scalability to match that of the server hardware they support.
Modular UPS systems with availabilities to 99.9999% are now possible
To see how a UPS system can best meet the requirements of its data centre users, we can look more closely at its function, and then at how it currently uses available technology to fulfil this task.
Incoming raw mains is fed to a rectifier/charger for conversion to a DC output. This output supplies the inverter input but is also used to charge the UPS battery (see Figure 1). When the incoming mains supply is available, the rectifier/charger keeps the battery fully charged, while the inverter also uses its DC level to develop an AC output for the critical load. If the AC mains supply fails, the inverter draws DC from the battery.
The battery is part of the DC bus, so switchover between battery and rectifier, and back again, is seamless. The mains failure is invisible to the critical load, provided it lasts less than the battery’s autonomy. The critical load is protected from incoming power aberrations as well as failures. The UPS rectifier and inverter provide a barrier to mains-borne noise and transient voltage excursions in addition to providing a well-regulated AC output.
Given that the UPS’s mains failure protection capability is subject to its battery autonomy, operators must have a strategy for handling power outages that exceed this autonomy. This strategy depends on whether or not the load must continue running throughout the mains failure. In some applications, it may be sufficient to ensure the ICT equipment has a safe, well-ordered shutdown. If so, a battery autonomy of around 10 minutes will usually be sufficient. This will allow the UPS, on detection of a potential extended power outage, to command the load to shut down within the battery’s autonomy.
However if the load must continue to run throughout the power outage, there are two further options. The first is to add batteries, increasing the autonomy time to hours rather than minutes. However a power outage event always has the potential to exceed the batteries’ autonomy, irrespective of its length. A more secure approach is to use the UPS in conjunction with a generator. On detecting an extended power failure, the UPS sends a start-up signal to the generator. The autonomy time then allows the generator to run up to speed and synchronise with the UPS’s AC output voltage waveform. With sufficient fuel, supply can be maintained throughout power outages of any length. The generator also recharges the UPS batteries.
Modular UPS topology
Early designs used a transformer to step up their inverter’s output to the required AC voltage level. Advances in power semiconductor technology, particularly the IGBT (insulated gate bipolar transistor) have allowed the transformer to be eliminated.
Transformerless UPS are about 5% more efficient than transformer-based ones. Efficiency improvement is sustained over the load spectrum from 100% down to 25% to achieve substantial reductions in electricity running costs and heating losses. Additionally, the power factor is improved, while total input current harmonic distortion (THDi) is reduced, bringing further cost savings and improved reliability.
While transformerless technology is extremely important for its energy savings, its reductions in size and weight also have far-reaching effects. These reductions result from eliminating both the transformer and the phase controlled rectifier. A transformer-based 120kVA UPS, for example, weighs 1200kg and has a footprint of 1.32m2. By contrast, a transformerless 120kVA UPS weighs just 310kg, with a footprint of 0.64m2.
This allows UPSs, even in high power installations, to be configured as sets of independent rack-mounted modules. For example, from one, up to five modules of 100kW each can be accommodated within a single UPS frame. A UPS can be scaled to a 100kW load with a single module, then incremented in 100kW steps to 500kW, matching the load as it grows. This flexibility in populating the frame is known as vertical scalability. For loads beyond 500kW, up to five more frames can be added, providing horizontal scalability up to 3MW.
UPSs can be configured as sets of independent rack-mounted modules
Alternatively, load up to 400kW load can be supported by five 100kW modules. This means that if one module fails, the other four can continue to fully support the load. As one module is redundant, this is known as N+1 redundancy. Modules can be ‘hot-swapped’; a process where a faulty module can be removed, simply by sliding it out of the UPS frame, and replaced with another without interrupting power to the critical load. This also has a positive effect on mean time to repair.
Modular UPS systems with availabilities to 99.9999% are now possible. They are equipped to fulfil the critical power protection role expected of them by today’s data centres – and they can do so with a true online efficiency exceeding 96%.
The last 30 years of UPS development has undoubtedly had a significant effect on IT power security and today, R&D is still responsible for driving growth, meaning further step changes in efficiency can be expected in the coming years. As the way data centres are used and managed develops over the next decade, UPS manufacturers will undoubtedly continue to invest and innovate, using the latest technological advances to ensure your load is as protected as it ever can be.
Figure 1; Incoming raw mains is fed to a rectifier/charger for conversion to a DC output for the inverter and the UPS battery charge.
Figure 2; Modular UPS systems with availabilities to 99.9999% are now possible.
Figure 3; UPSs can be configured as sets of independent rack-mounted modules.
Contact Details and Archive...
