Scientists at Harvard University have found a solution to their shortage of powerful computing resources by harnessing the Google Cloud platform to clone a supercomputer. The breakthrough could pave the way for other researchers to accelerate their work in the face of limited access to top-tier computing capabilities.
Professor Petros Koumoutsakos, who led the study, explained that their research focused on simulating a therapy to dissolve blood clots and tumor cells in the human circulatory system. This simulation required an enormous amount of computing power typically provided by a supercomputer. “The big problem that we had (was) we could run one simulation using a full-scale supercomputer,” Koumoutsakos said, noting that further refinement and optimization required continued access to the supercomputer.
In the United States, there are only a few supercomputers capable of performing the billions of calculations necessary to accurately mimic the conditions in Koumoutsakos’s study. This scarcity of resources has created bottlenecks in the scientific process, according to Costas Bekas, head of Citadel Securities research platform.
To overcome these bottlenecks, researchers and companies, including Citadel, have turned to the public cloud. However, cloud computing operations are not specifically designed to handle the demands of scientific research. Cloud systems are primarily built for reliability and resilience, with a focus on millions of relatively small computing tasks such as streaming video, serving webpages, or accessing databases.
Bill Magro, chief high-performance computing technologist at Google Cloud, explained that modifying cloud infrastructure to behave like a supercomputer requires changes in software, networking, and even physical design. Nevertheless, he sees the potential for cloud computing to unlock productivity and generate faster, better insights for scientific and engineering computing.
Koumoutsakos’s research was sponsored by Citadel and Alphabet Inc subsidiary Google. By leveraging the Google Cloud platform, the team was able to overcome the limitations of traditional supercomputer access and accelerate their study on dissolving blood clots and tumor cells in the human circulatory system.
This groundbreaking approach could inspire other researchers to follow suit and adopt the cloud platform’s capabilities in their own work. With the potential to eliminate the bottlenecks caused by limited access to supercomputers, this solution may revolutionize scientific research in various fields.
Max A. Cherney in San Francisco reporting. Sonali Paul editing.
Credit: The Star : Tech Feed