Shifting large computing jobs to nights or weekends could reduce their associated carbon emissions by a third, according to researchers who say that flexibility is key to reducing the environmental impact of the data centres that power everything from banks to social media networks.
The International Energy Agency found that data centres of power in 2019, or about 1 per cent of global electricity use, and says this is expected to rise to around 270 terrawatt hours in 2022. at the Technical University of Berlin says that in order to lower carbon emissions to aid with the fight against climate change, this energy use needs to be reduced, and could also be scheduled at times when renewable sources make up the largest amount of the energy mix.
Wiesner and his colleagues simulated the effect of changing the timing of large computing jobs to hours when the energy mix is skewed towards renewable power. In one set of tests, the researchers simulated a company running a nightly task every day for the year 2020. In the baseline test, used as a benchmark for comparing the findings, the task was scheduled to run each night at 1 am, but in other tests the task could be completed at any point during a 16-hour window. The model was given real-world historical energy mix information from each region during that year and told to minimise carbon emissions.
Advertisement
The simulation showed that in Great Britain, running data-intensive computing during this window would save 7.4 per cent of carbon emissions over the year. In Germany, the figure rose to 11.2 per cent, and in a simulation of California鈥檚 computing needs, the savings exceeded 33 per cent.
鈥淚f a scientist starts a machine-learning job on Thursday night, they don鈥檛 really care if this thing is done at Friday at eight in the evening or Monday morning,鈥� says Weisner. 鈥淪ay we have a nightly job, a nightly database backup, generating a financial report, something like that. And usually this would always take place at one in the morning. Maybe it doesn鈥檛 need to take place at one in the morning. Maybe it could take place anytime between six in the evening and eight in the morning.鈥�
Because the results of many large jobs aren鈥檛 needed immediately, there is scope to reduce carbon emissions with careful planning, says Wiesner. He adds that further work will be needed to determine how many current tasks carried out by large companies aren鈥檛 time-sensitive, but he expects that it will be a considerable ratio.
He says there is a precedent for a similar system 鈥� warehouse cooling systems and steel production factories often shift workloads to times when energy is cheapest. He says it would be possible to create government policy that incentivises a similar system designed to minimise carbon emissions instead.
at the University of Sussex, UK, says that 鈥渟uch solutions are exactly the type of thing that we need more of, particularly at the regional level, to move closer to net zero鈥�. But she says that it wouldn鈥檛 be a straightforward policy to implement, as working patterns may need to change and lots of diverse organisations would need to closely coordinate.
Reference:
