Selecting Server Rack Power Configurations

A datacentre is a power and cooling controlled environment in which servers are operated. There is no actual size-related definition and so even the smallest server room with a single rack cabinet can be a ‘micro’ datacentre.

The key point is that the power and cooling are controlled. Control infers management and some form of intelligent single-control loop decision-making systems such as air conditioners and uninterruptible power supplies. Within the environment all the systems should be connected through SNMP (simple network management protocol) to a software-based monitoring platform. Within larger datacentres, this would be a DCIM (data centre infrastructure management) package rather than individual ones for say UPS power and another for cooling elements.

To scale-up our ‘micro’ datacentre we then need to consider the components within the single server rack and how to provide them with critical power and cooling. A typical server rack will contain one or more servers and switches and be supplied by a PDU. Where the servers are dual-corded the PDUs will provide dual supplies. If the servers are three-phase powered, the input supply to the server cabinets will have to be a three-phase supply.

In a ‘micro’ datacentre running single-phase (230Vac 50Hz) server loads we would expect to see a single-phase UPS installed within the server rack. As the datacentre grows decisions then have to be made on how to power and cool additional racks.

It is not uncommon to find 3-4 racks with a small datacentre environment each with their own rackmount UPS and battery packs installed. This is especially so if the power draw for the entire rack is 3kVA or below. This is known in the power protection industry as a decentralized datacentre UPS system. Where all the server racks powered from a single datacentre UPS, the configuration is a ‘centralised UPS system’.

Traditional a centralised UPS system has always been the only option to supply critical power as the total power demand per server rack increases, and the number of racks increases. This made sense as it simplified the installation to a single datacentre UPS system and battery pack for power distribution, management and control, service and preventative maintenance. All you had to make sure was that the UPS was sized to cater for future expansion.

There is now a third way for datacentre operators and this comes via modular UPS systems. Modular UPS systems comprise building blocks and start with a single empty frame, similar to an unpopulated server rack. UPS modules that may run from 10-40kW are fitted inside the UPS frame to give a total capacity system or N+X configuration. The frame is sized to allow for future modules to be added at a later date.

For datacentres, the modular UPS system approach gives them an added option of being able to install the UPS system within the datacentre itself, as an in-row item. This means that a modular UPS system can sit between server cabinets or at the end of each row, in much the same way as an in-row cooling system. Its batteries may be housed in a matching cabinet or inside a combination UPS and battery cabinet or in a separate room. If the datacentre ambient is to rise above 25degrees Centigrade, the UPS batteries will be to installed in a grey space area or separate plant room.

The key with a centralised or modular UPS system within a datacentre environment is to forecast the final size, overestimate by about 20% for headroom and to have an electrical installation that can be scaled. The electrical cabling and distribution must be capable of supplying the future load demand or at least be capable of being economically expanded to do so.

Datacentres will standardize more and more on modular UPS systems. Whilst modular UPS do currently incur a 20-30% premium over traditional monoblock UPS, prices will fall over the next 2-5 years as more UPS manufacturers promote the solution and modular becomes the standard in datacentre UPS systems.