The Debrief: Robert Harrison’s 2,000X laptop

Robert Harrison: Start your computations.

As a chemist and a mathematician, Robert Harrison appreciates the vital role computers play in all modern sciences. As director of Stony Brook University’s Institute for Advanced Computational Science and chief scientist of Brookhaven National Laboratory’s Computational Science Initiative, he’s determined to make sure regional scientific and commercial interests appreciate it also. With SBU announcing the arrival of its new high-speed, $2 million SeaWulf computer cluster, the Cambridge University PhD and professor of applied mathematics has a fairly easy sell – even if he prefers you not call it a “supercomputer.” His words:

IT’S RELATIVELY SUPER: The price tag for a supercomputer these days is a ways north of $10 million. This is not that. On a national scale, we’re not at a supercomputing-resources level. But this is still an enormously important part of the university’s science strategy. And it’s very exciting. More and more science is affected by our ability to do faster computations and data analytics, so this is certainly super to us.

MUSTERING CLUSTERING: It’s 2,000 times faster than the typical laptop and has roughly 100,000 times more memory than a typical laptop and 20,000 times more disk space. Compared to what you can do with a desktop computer or your laptop, this cluster will have a transformative impact on science and research.

BUILT TO SPEC: We surveyed the science projects underway across campus and put together the proposal for the $1.4 million National Science Foundation grant that partially funded the new SeaWulf. From the outset, this was designed to meet the communal needs of the science portfolio happening here. One of the missions of the IACS is to pull these threads together. We want people from the outside to look at us and recognize this university is at the forefront of computational science.

STRETCHING THE DOLLARS: With the $2 million (including a $300,000 NYSTAR grant and a matching university stipend), we purchased four years of maintenance and upgrades from vendor Penguin Computing. The computing industry being what it is, I would anticipate that we’ll operate the system for up to five years, but within two or three years we’ll be acquiring some even larger resources. We have some pretty big plans for the university’s computational resources.

COME ONE, COME ALL: Whether it’s a researcher at Stony Brook or someone from elsewhere in industry or education, the uniform access mechanism is at the heart of what we do. The IACS has a fairly broad mission: Certainly, intellectual leadership is a big part of it, but economic impact is also part of the mission, including providing resources like this cluster for industries to use either directly or indirectly through collaboration with university faculty and students. We’re already working with a dozen regional companies on their product design, in their materials evaluation and with whatever else they need … obviously those people will be using this cluster, but we plan to open it also to industry across the state – and not just in the Long Island region, which is obviously our emphasis.

MOSTLY FREE: There’s a baseline of services with this cluster we’ll be able to provide for free. But we’re also establishing mechanisms whereby people who want more than a baseline can use it, and we have to figure out a way for them to pay for it. The goal is to create a tremendous value proposition for the researcher: Instead of having to buy your own cluster and manage it and pay for the maintenance and the space for it, you can just sort of buy a node on ours, or even a fraction of a node. That’s the model we’re trying to put in place.

AIN’T WHAT SHE USED TO BE: A lot of people are sad to see the first SeaWulf go, but really, the equipment is at the end of its life. We’ve already notified people the system will be shut down in December, and once we’ve backed up all the data it will be decommissioned. For security and privacy, the disks will be destroyed and the systems will be dismantled. It’s nine-year-old computing technology, so it really has zero computational value.

ON THE BRIGHT SIDE: We’re going to rescue all of the data, and most of the programs will smoothly transition from one (cluster) to the other, and they’ll immediately be running 50 times faster. People should like that.

BALANCING ACT: There is a clear delineation between my roles at Stony Brook and Brookhaven. But there’s also tremendous synergy. Both institutions want to expand computational science in the broader sense – not just because they want to compute, but because it’s essential to excellence in everything else they’re doing. There are a lot of common interests, which means we can executive a lot of strategy in our partnership. And there’s a commitment from both sides to create several joint appointments like mine over the next few years.

NOT HIS FIRST RODEO: I was the director of the Joint Institution for Computational Sciences, which sits between the Oak Ridge National Laboratory and the University of Tennessee. The relationship is very much the same as the relationship between Brookhaven Lab and Stony Brook University. Brookhaven is actually my fourth Department of Energy laboratory; I also worked for long stretches at Argonne National Laboratory (in Illinois) and Pacific Northwest National Laboratory (in Washington State).

DATA, DATA EVERYWHERE: People compute to understand climate change or to discover the Higgs boson or to try to figure out what’s happening with an ecosystem in Antarctica. As a chemist, I’m interested in helping people design new clean-energy sources. Computation cuts across everything. If you want to be excellent at anything, you need excellence in core computational science.

Interview by Gregory Zeller