| Data centre design and cost control |
|
|
|
| Written by Mike West, Managing Director Keysource 2009. | |
|
New initiatives are targeting the big issue of rising power consumption in critical facilities. Mike West explains. Energy efficiency is a topic rarely out of the public eye this far into the 21st century. Businesses and individuals are being encouraged to think about the impact they are having on the environment and to take responsible steps to conserve energy, and ultimately reduce greenhouse gas emissions. As upwards of two per cent of global CO2 emissions are estimated to be caused by information and communications technology, it is no surprise to see the data centre industry on the political and legislative agenda. Arguably even more significantly, industry has reached a tipping point where rising power costs can represent 25% of the total cost of data centre ownership - so businesses are challenging solution providers to create highly efficient facilities both to reduce carbon footprint and drive down costs, whilst also accommodating increasingly high-density hardware platforms without compromising resilience. There is, too, an increasing demand from businesses to gain a greater understanding of the performance of their data centre infrastructure and associated operating costs so that they can be managed more closely and more responsibly. In particular, businesses are looking for help and advice on achieving efficient power utilisation that doesn’t necessarily involve the construction of new facilities and that introduces more efficient cooling methods to tackle the increasingly powerful, high-density IT servers they are required to install. Efficiency measuresThe Green Grid is an industry-sponsored organisation dedicated to establishing metrics that measure the efficiency of data centres, and its members are qualified to advise on industry best practice and on guidelines to achieve the ultimate in operational energy efficiency. The key objective is to eventually establish one overall measure for data centre efficiency that includes the physical infrastructure (primarily power and cooling) as well as the IT layer (servers, storage, communication and networking). This measure could, in simple terms, be defined as ‘useful work over power consumption’. As a starting point, the data centre industry has decided to focus on DCiE (data centre infrastructure efficiency) and PUE (power usage effectiveness) which can be easily obtained by using standard user management tools. Adopting these standard measurements makes it easier for the industry to pursue three main energy efficiency goals: to minimise the overall power needs of the data centre; to maximise the percentage of the power coming in that is used for IT computing work; and to minimise the amount of power spent on non-IT computing equipment such as power conversion and cooling. Putting it into practiceAs an industry innovator, we ourselves have taken on the initiative of launching a data centre energy efficiency ‘benchmarking and measuring’ service that gives solid and sound advice on what’s needed for facilities to become more energy efficient. Given the limited data usually available to determine how well existing data centre facilities are performing, we commissioned a considerable sample number of initial surveys to establish the DCiE and PUE of UK facilities. However, it soon became evident that, whilst collecting the basic data was certainly beneficial for the overall exercise of measuring ‘efficiency’, we needed to understand why some facilities were performing better than others and how this information could be used to help clients fully understand what opportunities are available to them to improve performance within their own facilities. We therefore decided to benchmark key elements against ‘best practice’ to find out if observance of that practice definitely does result in energy reduction. We took a closer look at three sites with 60-70 racks with power consumption of between 231kW-270kW and conducted a seventeen point benchmark to best-practice each site based on construction, power, cooling, rack configuration and layout. What we found was that the best performing site supported more than double the IT equipment for a similar total power overhead of the least efficient site; and it was consistently operating within the best practice measures. Therefore, by adopting similar appropriate measures within the poorest site there was the potential to make significant energy and cost reductions that would support more than double the IT equipment; reduce power consumption by 122kW/hr with savings of £96K per year, and achieve a CO2 reduction of 36m2 or carbon 71kg/hr (621,960kg/year). Although the initial assessments provided an excellent start to the benchmarking exercise, it only represented a snapshot of the true picture and there are many variables that can affect the measurement. Collecting the data over a longer period of time, and on a continuous basis, will obviously provide a much clearer representation, and the initial measurement can now be used to track the facilities over time to understand and monitor the impact that energy efficient measures, equipment upgrades or new equipment deployment have had. As well as the realisation that benchmarking against other facilities is helpful in terms of establishing ‘best performers’, it became clear that data centres can be measured and categorised in many ways and this can have an enormous effect on the DCiE and PUE. As further measurement and analysis takes place, it will be possible to make further comparisons based on size, number of racks, rack power density, tier classification and redundancy levels, function and industry, utilisation, location and climate. Overall, as an initial energy efficiency benchmarking exercise for data centres it has provided a good foundation for identifying areas for improvement and comparison, which, in turn, can be used to give an indication of what’s actually needed to implement the energy efficiency improvements. Target your coolingThere have been many white papers published on this subject and it has become clear that ‘right sizing’ and the ‘level of redundancy’ have a considerable impact on the efficiency of data centre facilities, although this should not be taken as read for all facilities. However, what is most apparent is that it is the cooling function within the facility that provides the widest scope for energy reduction, particularly when applied through best practice and when concentrating on the fine-tuning of existing systems. Following the benchmarking survey, we were able to confirm an increasingly common industry realisation that considerable energy losses are also quite clearly due to poorly configured temperature and humidity controls and set points. It is vital that all the key elements that affect how heat is reduced and removed from the facility should be considered as a whole in order to achieve maximum efficiency. Hot and cold aisle layout with blanking plates is very effective; however, it’s the control of the warm return air and minimising hot and cold air mixing that is truly critical to avoiding hot spots, maintaining redundancy and reducing energy. There is no question then that the correct approach to energy reduction is to start measuring and establishing a baseline and track performance. The DCiE/PUE metric is a great starting point and easy to understand; however, to truly uncover the potential savings and to establish a robust energy reduction programme, it is essential to drill down into the infrastructure to get to the essential elements needed to make a proper assessment, taking into account ambient conditions and seasonal effect. More detailed analysis of the data centre space is clearly required, but progress is being made. The data centre industry needs to take the lead and be at the forefront of rolling out best practice energy efficiency assessments to businesses and incorporating these metrics into the design of new facilities. This will provide a road map to running a more efficient facility and ultimately result in reducing the energy consumption of a data centre - lowering the total cost of ownership and improving the overall performance of the facility. |
| < Prev | Next > |
|---|











