The Character of PUE

The simplicity of PUE has sparked an ill-informed rush to measure and compare the metric between data centres and organisations. In some cases the aim has been to demonstrate energy efficiency, in order to establish either technical credibility or marketing benefits. However, such comparisons have negligible value since PUE does not take any factors, other than TFP and ITEP, into account The metric does not acknowledge the different construction of each data centre, nor the variety of equipment, software, infrastructure, processes and other factors that make each facility unique.

Characteristics which can affect efficiency, but are excluded from the PUE metric include:

1. Occupancy
2. Workload
3. Technological Maturity
4. Location

Occupancy

The occupancy of a data centre can be considered as a measure of how full of equipment the facility is. Although there are several methods for defining this metric, for our purposes we will consider a 1600 m2 data centre that has only 400 m2 of its space occupied; therefore its occupancy is twenty-five per cent (25%). As more equipment is added to the facility, more floor space will be used, causing an occupancy increase.

Typically, a data centre facility comprises two major equipment types; the infrastructure required to operate the facility, and the IT equipment that is dependent on the infrastructure. In a traditional data centre design, infrastructure was originally scoped and installed to cover the facility’s expected life. This meant that the energy capacity and cooling capabilities of the data centre placed a theoretical maximum on the services that could be called upon. Unfortunately this approach did not lend itself to taking advantage of new technologies without significant cost or potential disruption.

However over time IT Equipment has become more energy efficient, with some facilities greatly extending their useful life.

The modern approach to data centre design is to implement a flexible and modular approach to the infrastructure. This has the advantage of delivering many benefits, including the ability to refresh the infrastructure as more efficient technologies come to market and an improved approach to purchasing costs on fixed assets and their reduced operating costs.

As such occupancy dramatically affects PUE and as discussed it is highly dependent upon the principles of design of the data centre.

As an example, if a data centre with a PUE of 2.2 is compared to another data centre in the same locality that measures a PUE of 3, it could be assumed that the data centre with the lower PUE is more efficient.

However, this conclusion cannot identify, and therefore overlooks, the fact that both data centres have completely different occupancy values, based on their configurations, usage or design.

If the data centre with a PUE of 2.2 is of traditional design with full occupancy, then it may be difficult to accommodate any advances in infrastructure technologies. Therefore, the PUE value of 2.2 is most likely to be reflective of its long-term PUE.

In contrast, the data centre with a PUE of 3 may have an occupancy of only forty per cent (40%) giving it considerable scope for expansion. As such its potential PUE at one hundred per cent (100%) occupancy could theoretically be 1.2.