Power distribution at the rack level


With a huge demand for computing devices, particularly enterprise servers that are generally housed in data centres or co-location facilities, and a commensurate rise in the use of power distribution units (PDU), organisations cannot afford to ignore their power use and environmental responsibilities, says Martin Hennessey.


Data centres alone consume approximately 3% of the world’s electricity and produce more than 200 million tonnes of CO2. These figures are continually growing and coupled with rising electricity prices, the costs of running a data centre can quickly spiral out of control.

The global data centre rack PDU market is primarily driven by the increased demand for data centres, many reports suggest that the global data centre power market will grow from $15.19bn in 2014 to $23.67bn by 2019, at a compound annual growth rate (CAGR) of 9.3%

Tier III: concurrently maintainable 

A Tier III data centre requires no shutdowns for equipment replacement and maintenance. A redundant delivery path for power and cooling is added to the redundant critical components of Tier II so that each and every component needed to support the IT processing environment can be shut down and maintained without impact on the IT operation.

Many business applications are managed through the cloud and server virtualisation. Increased usage of storage and network equipment makes the data centre environment more complex. Rack PDUs distribute power to this equipment to maintain application efficiency and performance.

As a result, data centre and IT efficiency is increasingly important as a strategy to reduce costs, manage capacity and promote environmental responsibility.

All of this demand leads owners and operators to having:

Major concern over energy consumption that is driving the need for both intelligent and non-intelligent PDUs at the equipment or data centre cabinet side

Concern over rising electricity costs and/or limited availability is further driving the need to measure, monitor and in fact switch power devices at the outlet level within the IT cabinet

Consolidation of hardware to combine environmental sensors and power distribution

Many of today’s intelligent rack PDUs offer the ability to simply plug in environmental sensors. The most typical sensors usually combine temperature and humidity allowing users to significantly reduce the number of required network data ports and IP addresses at the cabinet level while maintaining one central management solution.

It is typical for each PDU to be able to provide two ports for combined sensors, so for a user the cabinet could have up to four sensors per two redundant (A and B) feed PDUs.

Density and receptacles 

With the trend of the decreasing footprint of IT equipment and the virtualisation of servers commonplace, the intelligent rack PDU market has seen a swing towards far higher density units being installed at the cabinet level.

If sites are predominantly driven by single phase distributed supply it is usual to see up to four intelligent rack PDUs mounted per cabinet with a total of about 96 receptacles per cabinet – where sites have distributed three phase then it is common for the intelligent rack PDUs to use as many as 48 or more outlets or sockets per PDU.

The market has also seen that mixed outlet or receptacle types have become very common. This may allow for as many as three receptacle types per intelligent rack PDU (typically IEC C13, IEC C19 and UK) and is largely driven by new data centre sites that are receiving migrated or legacy equipment from existing or older technology locations.

Resiliency and redundancy  

The use of colours at the intelligent rack PDU and cord level has also become a market trend, with most organisations favouring different colours at the cabinet for the A and B or primary and redundant feeds into equipment. Most vendors will offer chassis in black, blue, red, white and gray as standard.

One of the simplest areas of providing a resiliency is reducing accidental cord removal at the IEC equipment cord level. Unfortunately, the international manufacturing tolerance between the IEC cord plug and the IEC socket receptacle allows in some cases for cords to appear loose in intelligent rack PDU sockets.

This can be resolved in two simple ways: either use an IEC locking socket or receptacle in the build/manufacture of the intelligent rack PDU; or deploy a very low cost solution of using an accidental friction release sleeve that simply fits over the cord plug largely removing the tolerance between the cord and the socket.

Given the mission critical nature of the environment, the intelligent rack PDU must be designed, built and manufactured to provide extremely high levels of resiliency – areas that can be used to benchmark this are hot swappable digital local touchscreen displays and hot swappable DC power modules that are typically used within the metered and outlet switched (WS) PDU and the outlet switched with outlet metering (WSI) PDU models (see box), and probably most importantly the use of latch able relays at the socket or receptacle level that will always supply AC power or are always on in the event of component failure.


With any rack mount device or in cabinet PDU one of the largest challenges is to integrate the device into the data centre environment without simply building up additional connectivity, cable run, patch panel etc. or network-related costs. Hence one of the features that is very popular is the ability to cascade at least 16 PDUs to one master IP Dongle or address. This in turn reduces cost as only one data network port per 16 PDUs is required and as most IT managers will tell you only one IP address needs to be issued and managed internally, this typically represents huge cost savings to organisations.

Using the IP dongle method, it is common for the vendor to supply free of charge software to manage at least 480 PDUs. Of course, with install densities on rack mount devices and intelligent PDUs rising all the time per site or across multiple sites, the management of the equipment/IP addresses required is typically being handled by other dedicated third-party DCIM platforms or an open industry interface known as SNMP.

It is evident an organisation cannot afford to ignore its power utilisation and environmental responsibilities.

Martin Hennessey has 25 years experience in the IT industry and is managing director EMEA at Austin Hughes.

Click here to see if you qualify for a free subscription to the print magazine, or to renew.

Follow us at @mcriticalpower. For regular bulletins, sign up for the free newsletter.


Please enter your comment!
Please enter your name here