One of the United Kingdom's fastest artificial intelligence supercomputers, known as Dawn, has been offline for more than a week after a recent heatwave caused it to shut down. The multi-million-pound supercomputer, owned by the University of Cambridge, was forced offline on June 27 when temperatures at the institution reached 30 degrees Celsius (86 degrees Fahrenheit).
Access to Dawn, which is used by scientists for research in fields such as climate and cancer, was expected to be restored on Monday. Last month was the hottest June on record for England and the second-warmest for the UK overall, according to provisional data from the Met Office. The UK's highest temperature for June was recorded on June 26 at 37.7 degrees Celsius (99.86 degrees Fahrenheit), surpassing a previous record set in 1957.
The shutdown occurred amid preparations for the country's third heatwave of the year, which prompted schools to cancel sports days and allow students to wear PE kits. The Met Office had predicted daily highs between 31C and 34C in London for the period.
A Cambridge spokesman stated that "technical issues during the hot weather" had caused the shutdown, but confirmed that full cooling capacity had since been restored. Staff at the West Cambridge data center, where Dawn is housed, were informed of a "major IT incident" resulting from a "failure" in the data center's cooling system. Supercomputers can be cooled via fans circulating air or through liquid-based systems.
Lennard Lee, director of the UK cancer vaccine and AI scientist and supercomputing project at the University of Oxford, indicated that research jobs were paused due to the outage but confirmed that no data was lost and no work would need to be repeated.
The Dawn supercomputer has supported research for the British Antarctic Survey in forecasting sea ice changes and has been used for climate modeling. It has also contributed to the development of personalized cancer vaccines, with Cambridge University noting its capability to detect cancer in kidney scans.
In January, the UK government announced an investment of £36 million to increase Dawn's processing power sixfold. The supercomputer had previously supported over 350 projects in the UK free of charge and was used for public initiatives, including efforts to reduce NHS waiting times. Its next iteration is planned to be known as Zenith.
Experts note that investing in robust cooling systems for supercomputers can be costly, with systems designed to withstand higher temperatures, up to 50C (122F), being significantly more expensive. Professor Simon McIntosh-Smith of the University of Bristol highlighted that the unavailability of such supercomputers poses a problem for the UK, as they are considered vital resources. He also suggested that more frequent extreme weather events could risk the performance of these machines.
The firm that supplied Dawn's cooling technology, USystems, a UK-based entity owned by French company Legrand, asserted that their systems operated within their design specifications and performed as intended during the event. Another expert, Jonathan Hirst from the University of Nottingham, commented that extreme heat appears to be occurring more frequently than anticipated and suggested that more efficient programming and data management could help data centers avoid overheating during heatwaves.
This incident follows similar shutdowns of data centers owned by Google and Oracle in July 2022 due to extreme heat.