Managing data centres in hot climates poses unique challenges. Can going green and using AI help?


By Jo-ann Huang
Let’s face it. On paper, Singapore is probably not the most ideal environment for managing data centres.
Given the country’s tropical weather, cooling one consumes a lot of power. In fact, electricity accounts for more than 50% of the operating expenditure of a typical data centre.
Put another way, the 10 largest data centre operators here consume energy that can match those utilised by about 130,000 Housing and Development Board (HDB) households.
But just because this is how it is currently does not mean it cannot change. The unforgiving weather is just one area that Singapore’s Infocomm Media Development Authority (IMDA) is tackling to green the nation’s data centres – and thereby maintain its lead as the regional leader for data centres.
IMDA is doing so via its Green Data Centre Programme, which aims to develop and deploy new technologies to make data centres more adaptable to harsh tropical climates.
Greening data centres is taking on urgency given Singapore’s embracing of a digital economy, with growing reliance on high performance computing (HPC). HPC is vital to a digital economy as it is able to run increasingly complex computer processes.

This was a hot topic at a panel discussion on next-generation data centres at Supercomputing Asia 2018 (SCAsia 2018), held on 29 March 2018.
“The big question is how do we achieve near-zero overheads when it comes to cooling Singapore’s data centres,” said panel member Professor Satoshi Matsuoka from the Tokyo Institute of Technology.
According to him, Singapore faces a unique challenge as data centres require large quantities of clean, treated water for cooling purposes. Clean water has to be used as wastewater could corrode or damage data centres due to possible proliferation of bacteria and algae.
An AI for data centres
Beyond having environmentally friendly data centres, tapping artificial intelligence (AI) to efficiently use energy in these facilities seems to be the next big thing.

Mr Wilson Ang, director of the IMDA’s Intelligent Computing Lab – who was also a panellist at the discussion – highlighted the agency’s 10-month study to apply AI to data centres.
“At the national level, we are taking a step and doing something about it. We are looking three to five years down the road, which is why we started the study,” he said.
The lab worked with the National Supercomputing Centre (NSCC) – which manages Singapore’s first national petascale facility with HPC resources – to research how AI can boost energy efficiency.
For the 10-month study, the lab also partnered with commercial data centre operators.
“In Singapore we are trying to improve our data centre power usage effectiveness average from 2.07 to hopefully below 1.2,” said Mr Ang.
It is important to improve energy efficiency as data centres are expected to consume 12% of the country’s total energy by 2020 – up from 7% currently.
“We can look at how to capture the knowledge of data centre operators through AI to speed up troubleshooting process and predict when the machine is going to fail,” he said.
“AI is a game changer. We see data centre operators beginning to apply AI but to get to a fully powered, self-managing data centre will take a while,” he concluded.