Case Study: Data Centre Energy Efficiency
“You cannot control or manage what you cannot measure”
This is a project funded under the JISC Greening ICT initiative. It involves monitoring power consumption in different parts of our University data centre and varying the metering topology during the project. Some metering is done at high granularity (measuring power consumption of individual devices and locations). In other cases, the metering is done at lower granularity (giving the combined power consumption of aggregated devices and locations from a single meter).
The accurate measurement of the power consumption of the three systems within the data centre (i.e. computing, mechanical and electrical) enables the calculation of the Power Usage Effectiveness (PUE) of the centre, which is also used as a comparative measure of energy efficiency for different parts of the centre. Another focus of the project is the temperature parameter and its effect on improving the energy consumption of the data centre using PUE.
Details of the methodology are found in these sources: Pattinson and Cross, 2013; Pattinson, et. al, 2015. At the outset of this project, new metering was installed in the Woodhouse Building’s LV switch panel and server room. The system schematic diagram of the metering topology is shown in the publications listed below.
All metered data for the Woodhouse Building is collated via the EGX300 gateway, which is configured to work with the web query function of Microsoft Office 2010. A spreadsheet is created that uses the web query command to auto update with real time information from the meters logging interval. When the spread sheet is opened it automatically connects to the EGX server and updates with logged data from the meters. PUE is a measure of how efficiently a data centre uses its power with respect to the amount of power that is actually used by the ICT equipment. It is a ratio of the total facility power to ICT equipment power.
Immediate impact of the research is:
The average Computer Room Air Handlers energy consumption seems to increase when the set point temperature is increased from 24°C to 26°C. However, the total ICT energy consumption seems to be constant, thus increasing the PUE.
A change in the type of cooling system (from Airedale AH DF units ((80kW Cooling Duty) with Trend OEM controls) to a new evaporative cooling system brings about a decrease of PUE from 1.7696 to 1.3262, which indicates increased energy efficiency.
The results from this experiment provide pointers to most effective changes in data centre configuration. However, the effects of changes in the set point temperatures would have to be explored further over a longer timespan and taking into consideration the different seasons of the year. This project also provides valuable insights into how a real data centre behaves in practice. Additionally, it shows the use of a range of benchmark techniques for comparison.
Recommendations for future work will be:
- extend the timeline for data collection;
- investigate the effects of virtualisation of the servers on PUE;
- include a wider range of parameter changes to the data centre configuration.