How effective is PUE as an industry standard for measuring data centre efficiency?

(This is our final blog in our PUE series)

When measuring efficiency in the data centre PUE gives us a starting point but, as with most standards, does not necessarily meet all the needs of the industry. Although we’ve touched on some of the limitations of PUE in our previous blog, it’s important that data centre specifiers and managers are aware of the challenges of using the PUE metric as an indicator of overall efficiency and are not mislead by some PUE claims.

Calculating the PUE metric takes time and effort and although the guidelines are very specific, it can be challenging. As we close off our series on PUE here’s our top five list of points to note when looking at PUE figures:

  1. Collection of Data

The time of day, the points at which the monitoring takes place, and the frequency that data is collected all influence the PUE results and make it very difficult to have an accurate comparable PUE indicator. There will be spikes in loads at different times in the day and even different times in the year. A simple example is that on the hottest day of the year more cooling is going to be needed than on the coldest day of the year. (See the full list of requirements on our first blog – What is PUE and Why is it Useful?)

While yearly results can help at an overview level, the ideal is to have continual daily monitoring to get a genuine impression of efficiency in a data centre. This can be quite an undertaking for a data centre manager and is more suitable to some environments than others.

  1. Partial PUE

Different data centre settings need different metrics. PUE is based on the premise that the data centre facility is only used as a data centre and does not factor in that other departments might also share space in the facility. It does not work well with the business model of data centre services or colocation providers either when they are not operating at maximum design capacity.

This has resulted in the Partial PUE metric that allows the active data centre zone to be measured instead of areas not being actively used for data centre services. This means that the partial PUE will indicate how efficient a data centre is at part or low load which is imperative to data centre services companies, (see our blog When is PUE not PUE).

  1. Miscalculating and Misreporting

Some managers find it difficult to understand the PUE metric and may report a PUE value of less than 1.0. This is impossible to achieve because all energy will have an output of at least 1.0 and therefore anything lower is not achievable.

Not all reporting is accurate because some of the aspects of the data centre have been excluded for example, calculations have been based only on the cooling system.  Or in the case of modular designs only the electricity supply within the modular environment has been measured rather than including the electricity points actually supplying the electricity to the modular unit.

  1. Non Comparable

Although PUE can provide an overview of how a data centre is performing over time, the complexity of the calculations and different data centre environments means that comparing the PUE across data centres is not very meaningful. To make useful comparisons, we need to look at the PUE calculations and reporting in detail.

  1. Limited scope

Efficient sourcing of energy is not included in the metric so energy-saving generation measures such as cogeneration, waste-heat re-use and local power generation are not reflected in the PUE value.

In conclusion, while PUE is a useful indicator, it is not a case of one-size fits all and analysis needs to be tailored so that it is relevant to the individual data centre environment. Correct monitoring and management, that are able to drive actionable insight are required to deliver real efficiency cost savings. You can find out more about our monitoring solutions here

When is PUE not PUE?

When Data Centre designers and builders talk about the efficiency of the facilities they have designed and/or built in terms of the industry standard PUE, they often reference the design PUE. We look at why this can be misleading and why it is important to understand the difference.

This blog is part of a series looking at PUE and data centre monitoring; you can read our first post, which looks at what PUE is and why it is important here.

Design PUE – What is it?

When a new facility is designed, a PUE can be calculated based on the energy consumption of the proposed plant, cooling etc. as a measure of how efficient the solution could be. So why is this a problem?

Well, the PUE figure being quoted is often based on the premise that the data centre is actually running at 100% IT load. This is often unrealistic, as most data centres, especially in the colocation market build up their loads over time, with it taking anything from three years or more to get near to operating at full-load. In reality traditional data centre efficiency rapidly deteriorates at lower IT loads (most facilities never operate at 100%). This means that the suggested power efficiency levels will not come in to play for a significant time, if ever, and operating costs will be substantially higher than indicated by the Design PUE.

So what should you do to address this?

Whilst Keysource is reasonably unique in the fact that we have in-house design, project management and operation teams (which allow us to constantly feedback and learn from past projects) we would expect most data centre design and build companies to understand the fact that most facilities need to be efficient not only at full load but also at part load. They should take time with you to fully understand not only the technical requirements of the facility but also your business, so they can explain and work through this with you.

When this is not done it can cause problems as, understandably, it can have big impacts on your operating costs. This limitation of the PUE calculation is of particular importance to companies that are working to tight margins to provide for example, cloud services and colocation services.

To keep operational costs under control, it is often best to consult with specialist data centre designers. They will be able to advise you on the best technologies to deploy that drive down part load operating costs from the start and who will be able to work with you to fully meet your business and technical requirements.

Sounds good, but how can this work in practice?

We recently completed a data centre for colocation specialist; ITPS (read the case study). Part of our process was to really understand their business objectives and use our experience within the regional colocation market to propose a solution that would deliver not only the highest efficiency but also allow high densities to be deployed anywhere within the data centre. Knowing that this efficiency was needed from day one the design was built to be highly efficient at part load. To further enhance this a modular solution was used to allow ITPS to deploy more cooling and plant as required rather than having it all in place on day one.

If you would like to understand more about PUE you can read our previous posts which explain the different ways PUE can be measured or download the top 10 things you need to know about PUE. As you may have guessed, it’s one of the areas we are passionate about so please call us and speak to one of the team.

We are happy to talk you through any of the issues and answer any questions about your current or planned facility’s PUE and, if required, give you some more information on how you can increase the efficiency of your facility.

Keysource Delivers New Modular Data Centre for Jaguar Land Rover

Keysource, the data centre and design build specialist, has announced that it has completed the prestigious data centre for Jaguar Land Rover’s new Engine Manufacturing Centre in Wolverhampton.

Due to the use of best in class modular design and Keysource’s expertise, the 1100 square feet data centre, which was commissioned in January 2015, has been successfully completed on time and within budget.

Andy Hayes, Director, Keysource explained,

Using a modular design means that the bulk of production and preparation can be done off site which reduces costs and saves time. Typically a project of this nature would take at least 12 months to complete but using modular designs reduced this by two thirds. Along with our cooling efficiency design measures, we see this modular approach as a template for future data centres.

As planned the core datacentre module houses 30 racks with a capacity of 320Kw. It provides critical power through an N+1 UPS system and 2N standby diesel generators. The power module designed and built off-site also offers scalability to support future expansion.

Cooling efficiency has been a key specification for this project and variable speed controlled fans have been positioned so that hot and cold air are separated via the wall rather than under the floor and this has resulted in the desired achievement of PUEL2YC of better than 1.2*

To PUE or not to PUE (That is the Question!) What is it and why is it useful?

For a few years now the data centre industry has been moving towards a standard measure of effectiveness for the use of power within data centres.  Power Usage Effectiveness (PUE), whilst not yet a globally agreed measure is fast becoming the standard and formalising the calculation is well underway. But what is it and why should you bother?

PUE allows organisations and enterprises to gather data, measure and report the effectiveness of their data centre, or indeed data room, in terms of power use.

In simple terms, PUE is,

Total Energy Used in the Facility
Energy Used by the IT Equipment

For smaller organisations or facilities that are located within a larger mixed-use space, Partial PUE (pPUE) can be used. This allows the data centre manager to measure PUE within a set boundary such as a room or building, or an area such as the equipment owned by certain customers or departments.  For example, you may want to measure the pPUE of individual data halls, rather than the facility as a whole.

The difficulty for many businesses is the availability of the data and information required. Whilst assumptions can be made and data captured at different points it does mean that comparing PUE or pPUE across market sectors is difficult. However the real value of PUE isn’t the creation of a global league table and winners awards, but more importantly PUE gives anyone with one or more data centres or data rooms the ability to measure and compare the effective use of power over time and across the business. (There are a wide range of intelligent monitoring solutions available for data centres, you can see some of the ones we offer by clicking on the link and keep an eye out for our upcoming blog post on the “5 things you need to know about DCIM”)

PUE gives everyone the ability to measure and improve the use of power in their data centres.

So what should be included in each group for the calculation?
Total Energy Used by the Facility

Plus the IT equipment listed below

Energy Used by the IT Equipment

Whilst the Total Energy Used by the Facility can often be gathered at source from the utility meter supplying the facility, or a meter just prior to the data room (for pPUE), the data collected regarding the Energy Used by the IT Equipment will be more accurate the closer the source of the information is to the individual units that consume the power. For example, gathering power usage from the installed UPS units is not as accurate as the information available from the subsequent Power Distribution Units (PDUs), which in turn is not as accurate as the data gathered from individual meters immediately prior to the IT Equipment.

PUE and pPUE account for this within the standard definition by allowing 3 variations, PUE1, 2 and 3, with PUE3 being the most accurate. There is also three different reporting frequencies; Yearly (Y), Monthly (M) or Weekly (W) with each denoting what period the data has been averaged over. On top of this you then have the data collection frequency; Monthly, Weekly, Daily or Continuously. So a PUEL3YC (which is the standard measurement we use at Keysource) would be a measurement from PDU level on a continuous basis. 

World Class PUE

Given the lack of consistency in measurements and environments it is not valid to quote a ‘world class PUE’, however it is generally accepted that a PUE of 2 or less is considered good and less than 1.4 is considered very good. A PUE of 1 means that 100% of the energy is used by the IT Equipment and therefore the physical data centre is 100% efficient, a PUE of 2 or less would mean that 50% or more of the power used by the data centre is used by the IT Equipment and so on.

There are so many factors that effect the efficiency of data centres that “snap shot” PUE figures can be misleading.  For example:

Because of this, continuous reporting and monitoring is the recommended way of tracking PUE, by not only Keysource but also The Green Grid; who developed the metric in the first place.

As touched upon it is also important to take into account the different environmental conditions of different regions. As such, PUE levels cannot be compared across regions because a higher PUE in one region might be relatively better than a PUE in another region but we will look at this in more detail in another post.

Remember, the key is consistency, define your measure, data sources and frequencies and stick with them. 

Design PUE or the PUE of new facilities

When a customer wanting a new facility designed and/or built approaches Keysource, they often have an idea of the PUE they want to achieve and a design is developed to meet this business need whilst also meeting the technical requirements.  However as this is a theoretical figure (admittedly based on some very complex mathematics!), how can you, as the customer, be sure that this is what you will be able to achieve and what if anything should you be aware of?  Find out more in our next blog looking at PUE.

[x]

Contact Us

Please fill in the form below and a member of the team will be in touch shortly.