A supercomputer in the cloud, built in just 33 minutes and capable of running a drug discovery application in mere minutes, has been announced by its creators YellowDog, a UK cloud workload management company.
The move looks to reduce the cost of high performance computing and make it more accessible, allowing critical innovation in fields such as engineering to accelerate from months to hours, all with less of an environmental impact than existing physical supercomputers.
In sixty-five minutes YellowDog’s cloud-based system – with power equivalent to that of the world’s sixth largest physical supercomputer – ran a popular drug discovery application on 3.2 million virtual processor cores, as a single cluster. To provision one million pre-emptible instances took just 7 minutes.
The technology not only provisions the exact resources needed, when they’re needed, but it also enables the selection of data centres based on renewable energy generation. This means high performance computing becomes much more efficient and computing tasks can be placed where clean energy is most available. This capability to move to where the sun is the brightest, or the wind the strongest, has never before been achieved.
“This number of concurrent cores, stretching across the Atlantic, represents a huge breakthrough. It reduces the time, cost and, so importantly, carbon impact of research into some of the most important societal issues,” said YellowDog CEO Simon Ponsford.
“Our platform, through its ease of use, lets organisations focus on the problem itself, not on being an expert in exploiting the potential of the cloud. The realms of possibility are no longer bound by coordinating computing resources, but by the vision of the human mind.”
“In the past year alone, we have seen not only the continuation of the global pandemic, but also significant displacement caused by extreme weather events,” continued Ponsford. “The ability to solve these complex and multifaceted issues hinges on collecting, processing, and generating insight from massive data sets. This demonstrates the growing importance of high performance computing in the efforts to advance the fields of science, engineering, and technology.”
In under 20 years, storing data is set to create 14 per cent of the world’s emissions – roughly the same proportion as the US’s total output today, with data centres use enormous amounts of energy, with cooling equipment accounting for some 40% of that usage.