What is a high-density server rack?
Rack density is measured by the amount of heat generated and power required to run all of the devices in a network or server rack. Typically for 42U network & server racks at the edge of the network ,the rack density would range from anywhere between 2-5kW. Rack applications with densities around 8-10kW would currently be considered high density racks and would usually host a high number of servers rather than just a couple of network switches, all of which can be support by a single phase rack-mounted UPS.
However, the emergence of high-performance computing (HPC) required to support AI workloads is rapidly changing what is now considered “high-density”. AI applications run on powerful GPU servers that demand up to 10 times more power than a standard CPU server. The result is a massive increase in the rack density for AI workloads and delivering this power to the rack presents some new challenges.
Rack mounted UPS simply aren’t up to the task of supplying a rack or multiple racks with densities as high as 100kW per rack. That means alternative methods of providing back up power need to be considered.
Deploying larger 3 phase UPS such as the APC Galaxy from Schneider Electric, or Riello MultiPower that are positioned in the row or in a dedicated UPS room, delivers higher capacity back up power and their modular UPS design provides the high levels of redundancy required to ensure the maximum uptime and uninterrupted service these applications demand.
Distribution of that power is also a challenge. Common 16amp and 32amp single phase rack PDU’s no longer have the capacity required to supply high density server racks with the amount of power they need. A shift to higher rated 63amp 3 phase and even higher rated rack PDU’s is expected, with manufacturers like APC by Schneider Electric already planning to release new rack PDU models specifically to meet this demand.
Inevitably where there is power being consumed, heat will be generated! High rack densities above 20kW can’t be cooled using traditional air cooling technologies alone and with the growth in deployment GPU servers to support AI workloads, cooling methods in data centres and edge computer rooms will need to change.
Liquid cooling technologies already exist to handle high density cooling requirements and whether direct to chip or immersion, they will become a standard part of a data centres critical infrastructure in the not too distant future.
If you’re working with high-performance computing or planning to deploy AI server applications, get in touch with our team today to discuss how we can help.
For more information about power, cooling and racks for AI applications or to request a quotation, get in touch with Source UPS today!
Call: 01252 692559 or email: info@sourceups.co.uk
We have worked with Source UPS for several years now. We have a large number of rack locations throughout our busy campus, but the service visits are always carried out with out any problems. The team are very accommodating and have made some useful recommendations that we have adopted across our campus network. I would definitely recommend Source UPS.
