Nov. 18, 2020 — In an educational environment where cutting-edge research involves increasingly staggering amounts of data, New York University today unveiled its new supercomputer, Greene, which balances the need for next-level computation and storage capacity with the need to reduce reliance on fossil-fuel consumption. The new high performance computing (HPC) cluster will bolster research across a wide range of disciplines, from biomolecular genetics to the political ramifications of social media behavior to artificial intelligence.
The new Greene supercomputer, built by Lenovo, has the capability to do four quadrillion (4 x 1015) calculations per second, making it 10 times faster than both NYU’s current supercomputer and the most powerful computer in the New York area. Greene connects to NYU research facilities with the fastest network available in any institute of higher education in the United States, and is one of the top 10 most powerful supercomputers in US Higher Education, according to the November 2020 Top500 supercomputer ranking.
“High-performance computing, big data and artificial intelligence are critical in so many areas of research throughout higher education and in particular here at NYU,” said Andrew Hamilton, President of New York University. “It’s of paramount importance that we continue to provide our researchers across the disciplinary spectrum with the resources they need to expand knowledge beyond what we may not even be able to conceive of as yet. But for the sake of our planet, we must also find ways to maintain this computing trajectory while simultaneously moving away from our reliance on fossil-fuel consumption. Greene is a wonderful example of how we can achieve both these goals.”
“Cutting-edge research throughout the university absolutely requires next-level computing power if we’re to continue to foster major breakthroughs, as well as attract the most talented faculty and students,” said Stacie Bloom, Vice Provost of Research at New York University. “Currently, researchers have to wait in a queue for computing power. Greene will have 10 times the capacity of the former cluster, with all computing for overflow moving to the cloud, both eliminating the bottleneck and making new and exciting avenues of inquiry possible.”
More Sustainable Supercomputing: Despite its increase in computing power, Greene will ultimately consume less power overall, incur lower usage costs, and be more sustainable in the long term than its predecessor, due to the following innovations:
Greater Efficiency: Greene will have a power usage effectiveness (PUE) rate of 1.35 or less, compared with the PUE of 2 for the current NYU “Prince” current computer. (A perfect PUE is 1.)
Hydrocooling: Greene will have water-cooled compute nodes, enabling portions of the computer to achieve a 1.08 PUE. For 30,000 CPU cores, the water will remove 95% of the heat, without any air conditioning.
More Efficient Data Center Equipment Arrangements: White racks will reduce heat consumption by 15-20%. The creation of a “hot aisle” (where computers are arranged back to back) will enable more efficient heat trapping and ventilation through the roof.
“We’ve worked very closely with Lenovo and EYP on the design and installation of Greene and the results have been phenomenal on all fronts,” said David Ackerman, Chief Digital Officer and Associate VP for Research Technology. “We have the most powerful university-funded supercomputer in the country, operating on one-third of the cost of the power of its predecessor. Innovations like hydrocooling and ‘hot aisle containment’ targeted ventilation ensure we can keep expanding our research capability while helping to achieve our institutional goal of reaching carbon neutrality by 2050.”
The Greene HPC cluster was the result of a collaborative effort on the part of researchers and administrators from NYU’s Courant Institute, Faculty of Arts and Science, and the Tandon School of Engineering, and the NYU IT Research Technology Division, which provides computing resources, services, and expertise for researchers from across the university. Major areas of research that have benefited from NYU’s next level computing power include:
- Artificial Intelligence: such as research by Yann LeCun, Turing Prize winner and Silver Professor at NYU’s Courant Institute, in computer vision, mobile robotics, bio-informatics, biological image analysis, medical signal processing, and financial prediction;
- Virtual Reality: including Courant Professor of Computer Science Ken Perlin’s work generating award-winning innovations in graphics, animation, new virtual reality, and augmented reality experiences;
- Genomics: including genome reading techniques for rice developed by Professor of Biology Michael Purugganan to develop a sturdier strain of rice that could address the ongoing issue of world hunger;
- Climate Modeling: including Professor of Mathematics and Atmosphere-Ocean Science Olivier Pauluis’ research on modeling tropical cyclones and the global hydrological cycle; and
- Computational Chemistry and COVID-19 Treatment: including Professor of Chemistry and Mathematics Tamar Schlick’s work to better understand the structural biophysics of the SARS-CoV-2 virus and the molecular mechanisms the virus uses to infect humans.
“We have solutions to many of the world’s problems sitting in the data sets we already have,” said Richard Bonneau, Professor of Biology,Computer Science, and Data Science. “Cracking those codes and finding those causal relationships is something that we’re really excited about trying, and it’s something that we can do now that we have this computing capacity readily available.”
“You cannot really do good AI research today unless you have access to a supercomputer with a large number of GPUs, and Greene has the kind of performance you need for AI research when you need it,” said LeCun, a professor at NYU’s Courant Institute of Mathematical Sciences and Center for Data Science and vice president and chief AI scientist at Facebook.
“Greene provides great opportunities for our students; we’re able to get them working on the HPC, giving them access to and practice with the same resources that our most advanced researchers are using,” said Andrea Jones-Rooy, director of Undergraduate Studies at the NYU Center for Data Science. “The fact that we can give our students access to this means they can already engage in cutting-edge research while they’re still pursuing their degrees, at both the undergraduate graduate and graduate levels.”
The NYU Greene HPC cluster consists of 665 servers and 32,000 CPU cores, and it is equipped with 332 GPU cards. The total memory on the cluster is 145TB and comes with 9 PetaBytes of GPFS data storage.