3D printed clouds and figurines are seen in front of the Google Cloud service logo in this illustration taken February 8, 2022. REUTERS/Dado Ruvic/Illustration
(Reuters) - A scientist at Harvard used Google's cloud platform to clone a supercomputer for a heart disease study, in a novel move that other researchers could follow to get around a shortage of powerful computing resources and speed up their work.
The study simulated a therapy that aims to dissolve blood clots and tumor cells in the human circulatory system that required an enormous amount of computing power that typically can be harnessed with a supercomputer, according to Harvard professor Petros Koumoutsakos.
"The big problem that we had (was) we could run one simulation using a full scale supercomputer," Koumoutsakos said, adding that refining or optimizing the simulation required further access to the supercomputer.
In the U.S., there are only a handful of supercomputers that are capable of running the billions of calculations to accurately mimic the conditions in Koumoutsakos's study.
The small number of machines capable of performing the research has created bottlenecks in the scientific process, according to Citadel Securities research platform head Costas Bekas.
To eliminate the bottlenecks researchers and companies such as Citadel that need an enormous amount of computing resources only found in supercomputers have begun to turn to the public cloud.
But cloud computing operations aren't designed to handle the demands researchers have. They are designed for millions of individual, relatively small computing tasks - things such as streaming video, serving webpages or database access. The cloud is usually built for reliability and resilience.
"Folks are realizing the potential for cloud to solve problems and technical scientific engineering computing to really unlock productivity and get to better answers, better insights, faster," said Bill Magro chief high performance computing technologist at Google Cloud.
Modifying cloud infrastructure to behave like a supercomputer requires changes in the software, networking and physical design of the hardware, Magro said.
Citadel helped sponsor Koumoutsakos's research with Alphabet Inc subsidiary Google.
(Reporting by Max A. Cherney in San Francisco; Editing by Sonali Paul)
