Why Fugaku, Japan’s fastest supercomputer, went virtual on AWS

Mount Fuji in Japan.

When you think of climbing one of the world’s most famous mountain peaks, ease of access isn’t the first thing that comes to mind. While Mount Fuji in Japan is famous for its height and width, it is also reachable by novice hikers without lots of time on their hands due to the nation’s efforts to make it accessible. Now, the researchers behind one of the world’s fastest supercomputers, Fugaku, are trying to make the supercomputer just as accessible on the Amazon Web Services (AWS) Cloud. Fugaku, a name familiar to many in Japan, is the supercomputer developed by Japan’s RIKEN Center for Computational Science (R-CCS) in collaboration with Fujitsu, as a result of the Japanese government’s national policy.

“Fugaku is another name for Mount Fuji. We named our supercomputer Fugaku because Mount Fuji has the highest peak in Japan. But it also has a very broad base, just like how we want more people to use Fugaku for broader societal issues,” said Dr. Satoshi Matsuoka, the director of R-CCS.

From basic science to real-life impact

Supercomputers have one million times more compute power than the fastest laptop. Fugaku is a massive supercomputer with 160,000 nodes and 8 million central processing unit (CPU) cores. Since its completion and the start of full-scale operations in March 2021, Fugaku has been used by numerous fields in response to social needs.

During the COVID-19 pandemic, Fugaku was used to perform detailed simulations of how the virus dispersed when exhaled. Other examples include searching for drug candidates, forecasting of economic and social impact, as well as next-generation solar cells, offshore wind power generation, linear precipitation zone forecasting, heart simulators, and many more.

Super leadership behind the supercomputer

Dr. Matsuoka at AWS re:Invent 2023, where he held a session on Virtual Fugaku, a replication  of the original environment on AWS.

Dr. Matsuoka at AWS re:Invent 2023, where he held a session on Virtual Fugaku, a replication of the original environment on AWS.

We can’t talk about Fugaku without talking about Dr. Matsuoka. In April 2018, he was appointed as the director of R-CCS, where he oversaw the supercomputer’s development. He is the trailblazer behind the democratization of Fugaku and the proliferation of its real-world applications for endless societal challenges.

“Mount Fuji is quite easy to climb despite its high altitude because of the well-maintained roads and trails for everyone. That’s the ideal representation for Fugaku: the easiest supercomputer to use in the world. Fugaku will make the innovation cycle much faster,” said Dr. Matsuoka.

Going virtual with AWS

Doing science on the cloud takes leadership. When Fugaku was running on premises, it was “at 95 percent capacity and constantly full,” said Dr. Matsuoka. With the AWS Cloud, Fugaku can leverage greater compute capability and impact.

“For these scientific outcomes to proliferate and create an actual impact on society, just building this 160,000-node machine is far from sufficient. The value of Fugaku is not only the hardware, but also the software solutions including the applications. That’s why we worked with AWS to extend our assets over to the AWS Cloud,” said Dr. Matsuoka. “We’ve injected this supercomputing environment directly into the cloud to extend applications and have much larger impact than in a single machine.”

The collaboration with AWS led to the creation of Virtual Fugaku, replicating the original environment on AWS. Virtual Fugaku provides a way to run software developed for the supercomputer, unchanged, on the AWS Cloud.

Virtual Fugaku on AWS can be easily used for cutting-edge research that companies wish to keep confidential, thereby expanding the range of its use. For example, GENESIS, a molecular dynamics software used in drug discovery and materials development, became available on AWS in December 2023. Notably, Virtual Fugaku on AWS proved to be twice as fast as the on-premises counterpart in complex calculations and simulations during proofs of concept (POCs) conducted in 2023.

Paving the way for scientific workflows

“Science today is a collection of various components in the workflow. You really need a platform to allow scientists to easily build their customized workflows. And for that, AWS is essential in science infrastructure,” said Dr. Matsuoka. “These hyperscale nodes will also advance AWS’s abilities, and there are a lot of things we can do together for a faster cycle of innovation,” he added.

Fugaku is not just a supercomputer; it’s a catalyst for scientific breakthroughs, bringing the power of computation to the forefront of societal advancement. And the journey continues.

Additional reading

Learn how RMIT University researchers leveraged an AWS-based HPC platform to visualize molecular structure 100 times faster.