The National Center for Supercomputing Applications and the University of Illinois Urbana-Champaign plan to decommission the Blue Waters supercomputer by the end of the year.
Launched in early 2013, the system was capable of 13.3 petaflops of peak performance. Blue Waters was the fastest supercomputer at a university anywhere in the world until the Frontera supercomputer at the Texas Advanced Computing Center came online in 2019.
NCSA is currently deploying its next supercomputer, Delta, for early 2022.
Blue Water was originally planned to come online in 2011, funded by a $208 million grant from the National Science Foundation (NSF). But the IBM Power7-based system had to be reworked when the deal with IBM was scrapped due to “increased financial and technical support by IBM beyond its original expectations.”
Cray (now a part of HPE) was brought in later that year to save the project, redeveloping it as a 281-rack supercomputer with 49,000 AMD CPUs.
It launched in 2013 at the National Petascale Computing Facility at the University of Illinois at Urbana-Champaign, itself built with $65 million from the state of Illinois. The 88,000 square foot (8,200 sq m) data center is home to other NCSA supercomputers.
The NSF then gave the university and NCSA $150 million to support the supercomputer's operations for five years, but the university was able to run it for less than expected, keeping operations going for a few more years.
In 2019, the National Geospatial-Intelligence Agency provided funding to cover Blue Waters' operations for another two years, with the contract ending at the end of this month.
During its nine-year run, Blue Waters helped create digital elevation maps of the entire Earth, helped understand black holes and gravity waves, and studied viruses such as Covid, HIV, Zika, and influenza.
It was also used by University of Illinois researcher Wendy Cho to study gerrymandering in Ohio, with the results included in a landmark gerrymandering decision from the Ohio Supreme Court ruling their congressional districting was unconstitutional.
“Blue Waters enabled computational, data analysis, and machine learning/artificial intelligence investigations that could not be done otherwise," William Kramer, PI, Blue Waters Program Director, NCSA, said.
"Through the years, the increasing numbers of projects identified as 'data intensive' processing or as doing machine learning/artificial intelligence showed the leadership of Blue Waters and the impact it was making. Many projects were not just tightly coupled simulations, data intensive analysis or machine learning, but also combined these methods with multi-scale and multiphysics simulation to achieve cutting-edge results."
Since Blue Waters went into production on March 28, 2013, it has provided 41.8 billion core-hours to scientists and engineers across the US.