Specifying data center pod architectures - Schneider Electric's latest White Paper
Schneider Electric has released a new white paper that lists the factors that need to be considered when specifying data center pod architectures.
The latest White Paper from Schneider Electric, a company that specialises in energy management and automation, explains how to specify the physical infrastructure for an IT Pod.
The report finds that the standardisation of IT equipment and support infrastructure at rack level helps to simplify the process of scaling up data centers to meet increasing load demands.
The paper also states that fully integrated racks loaded with IT, that can be rolled into place, provide a familiar way of upgrading data center resource quickly.
For centralised data centers, including the latest hyperscale facilities, there is a need to develop larger increments of IT resource, known as Pods, comprising a group of racks in one or two rows to facilitate rapid upscaling.
But, currently, no equivalent industry standards exist for such deployments, meaning operators must design and specify their own Pod architectures, claims Schneider Electric.
The report, Specifying Data Center IT Pod Architectures, which is White Paper No 260 from Schneider Electric, outlines the optimum configurations based on available power feeds, physical space and average rack power densities that should be considered when designing an IT Pod.The advantages of IT Pods:
For hyperscale data centers, in addition to the convenience of being able to expand using larger increments, a Pod can also be used as a logical grouping of business applications and can be assigned in entirety to a single significant client or line of business.
As the report states, Pods can be used to vary the technologies available in a large data center, for example housing Open Compute Project (OCP) racks in one Pod and traditional server racks in another.
The report goes on to explain that electrical redundancy can be varied Pod by Pod so that critical high-availability applications requiring dual power feeds can be kept separate from less critical ones, thereby maximising investment where it is needed and reducing cost where it is not.
Schneider Electric outlines the three main drivers determining Pod architecture:
- The choice of electrical feed
- The physical space available, i.e. the number of racks, for a Pod
- The average rack density required
The White Paper suggests guidelines for the power requirements of a Pod, recommending that each should be treated as either a low-power assembly running at 150kW or a high-power version capable of 250kW.
Schneider Electric explains that grouping racks together into Pods, with each Pod having a dedicated electrical feed, helps avoid the problem of complex power distribution that often emerges in a data center when some racks must "borrow" breaker space from Power Distribution Units (PDUs) that are not physically close.
To utilise physical space in the data center, meaning the longest practical Pod for the room, the design should balance power with rack density.
But, as always, with each individual, IT room specific challenges arise, which must be considered. These challenges could include the shape of the room, building columns and ducting.
When determining rack density, the paper recommends that designers and operators should underestimate the expected rack density because it is more expensive to deploy IT below the data center design density than to deploy above the design density.
To conclude, the report explains that standardising Pod designs and limiting the number of configurations can make Pod level deployments quicker and easier. Organising IT racks into Pods makes it easier to vary power and cooling redundancies and architectures based on specific business needs.
Robert Bunger and Patrick Donovan are the authors of the Schneider Electric White Paper, Specifying Data Center IT Pod Architectures.
Bunger is the director of Data Center Industry Alliances for Schneider Electric's IT Division and Donovan is a senior research analyst for the Data Center Science Center at Schneider Electric.