Power-hungry data centres to go on permanent diet
EU-funded researchers are on the verge of slashing the energy used by internet data centres to help cut CO2 emissions. By developing a model to optimise energy in real time, they aim to make the centres more efficient without affecting performance.
© MetinSeven #268108438, source:stock.adobe.com 2019
When Apple decided to open its first data centre in the European Union, it looked at opportunities in Ireland. However, the company dropped its plans after an environmental impact assessment estimated a 6 to 8 % spike in Irelands national electric power grid.
Apple eventually opted for Denmark, but the broader issue of such centres and their intense energy consumption remains a huge problem.
The EU-funded GREENDC project is targeting internet data centres, also known as IDCs, precisely because they are so energy intensive.
Due to the huge amount of energy implied and the related costs, IDC can make a significant contribution to energy efficiency by reducing the energy consumption and power management of IT ecosystems, says GREENDC project coordinator Habin Lee of Brunel University in the United Kingdom.
In 2017, a study by the European Commission found that the information and communication technology sector, including data centres, generates up to 2 % of the global CO2 emissions.
In the United States, IDCs alone used 61 billion kWh in 2006 or about 1.5 % of US energy consumption. US government estimates it will increase to 140 billion kWh annually by 2020 as more centres are built. A lot of that energy goes into keeping the centres cool.
In fact, Lee says that IT and cooling devices consume up to 90 % of the energy needed to run a data centre. Keeping the devices cooled is required to make sure they run properly, but getting the right mix between the two for optimum performance can prove difficult.
While most researchers simply seek to reduce power consumption of IDC, GREENDC has taken a novel route by focusing on improving the efficiency of how the centres operations run.
As part of that objective, the project researchers are using one of Turkeys largest data centres to test a simulation model. They have placed sensors on the operating devices at the data centre to measure temperature changes in real time between heat generated from servers to handle workloads and cold air produced by installed air-conditioners.
Using the real-time data collected at various places throughout the data centre has helped us to understand the individual impacts of air-conditioners on the inflow temperature of the servers, Lee explains.
The big-data analytics approach behind the project means they can derive a statistical model which explains how much each air-conditioning unit affects the servers inflow.
Twenty percent leaner
Given that the project is still ongoing, GREENDC plans to run a field test in March 2020 in the Turkish data centre. Lee says that the costs of energy to run the centre currently hover around EUR 1 million.
Our approach is expected to reduce the cost by 20 % or more, he says. Should everything go to plan, GREENDCs big-data analytics approach to optimising efficiency can be rolled out to other data centres through a web-based decision-support system developed in close cooperation with the Bulgarian SME which is also part of the GREENDC team.
Indeed, GREENDC consortium partner Turksat is going to use the decision-support system once the field trial has been successfully completed, Lee confirms.
Companies are not the only beneficiaries of this international research. The project is helping to train several PhD students through exchanges with industrial partners and training programmes. The experience gathered in this project is contributing to the career development of young researchers, as was the case of one student who has completed a PhD and now works for the United Nations on sustainable issues in Africa.