‘Things you couldn’t do otherwise’ | National Center for Supercomputing Applications at the University of Illinois
‘Things you couldn’t do otherwise’
09.09.11 - Permalink
After spending decades as a particle physicist, the University of Illinois' Jon Thaler now spends the bulk of his time on two massive astronomy projects that NCSA is also heavily involved inthe Large Synoptic Survey Telescope and the Dark Energy Survey. These instruments will start scanning the night sky in the coming years, allowing a host of new insights in astronomy and cosmology. They'll also produce amounts of data that are unheard of in optical astronomy.
Thaler talked to J. William Bell about the relationships between physics and astronomy, between NCSA and other parts of the University, and the importance of both.
Q. From particle physics to astronomy, how do you go from one to the other?
A. Life is too long for one career. I did conventional particle physics forever. Around 2002, I was chairman of our [department's] colloquium committee, and one of the people we invited was Tony Tyson. He's the head of the LSST [Large Synoptic Survey Telescope] project. We were going to lunch and he described what a big astronomy sky survey can do, in particular on dark matter. And I said, "What can a particle physicist contribute to a project like this?"
He said, "Data acquisition." It turns out that's what our group is particularly strong at. It meant we were able to make technical contributions to the project while we were learning the science.
Q. Particle accelerators produce huge amounts of data. Observational astronomy produces huge amounts of data. That's the connection?
A. Well, observational astronomy is going to. That's the change. The newer generation of astronomical surveysstarting with the Sloan Digital Sky Survey, which NCSA is involved with, and LSST, which NCSA is also involved withthey produce much more data than the traditional optical telescope.
For example, for the LSST, each image is 6 gigabytes of data, and we take a picture every 15 seconds. That's vastly higher data rates than astronomers are used to, but for the particle physicists, that's everyday life.
A typical particle physics experiment actually does quite a bit of what is called real-time data processing.
Q. And that's less the model in astronomy?
A. So far. The optical astronomers, you take a picture, it's sent. Now that's assuming that the experiment, the apparatus, the camera is working properly. So the thing you have to do in real time is what is called quality analysis or image health or instrument health. The important goal of this analysis is to avoid wasting a night taking poor quality data.
The LSST...costs $40 million a year to run. So you're talking about a lot of money per day. It's worth putting some effort into diagnostics and monitoring.
Q. So the data is packaged and then, in the case of LSST, it will be sent to NCSA. Give us a sense of what that part of it is.
A. There will be some specialized pipelines that are geared at analyzing the data for a particular science goal. For example, there's a supernova analysis.
In these surveys, you're taking a picture of a particular spot in the sky every few days. Most of the time, there are no supernovas. But occasionally an explosion will go off and there will be a new spot of light. So you need an analysis that asks, "Oh here's the picture I took tonight. Here's the picture I took three days ago. Is there anything different?"
The things that make it complicated are, well, the image of a star on one night doesn't look the same as the image of the star on another night because the quality of the atmosphere will change. The spot will be a little fuzzier or less fuzzy, and we have no control over that 'cause it's weather.
Q. And that's one of these data analyses going on at the far end before intensive analyses by individual research teams?
A. That's right. The idea is, now they say, "Oh, it looks like we have a supernova here." Well, all we really know is that a flashbulb went off. It could be anything. So then a person has to look at it in more detail. Some of this detailed analysis requires other observations with other kinds of instruments, and it has to be done while the supernova is still bright enough to see.
One of the important features of LSST is that it will generate alerts so people can look at time-dependent phenomena. Alert follow up is pretty much someone else's job...it's not the job of LSST itself to do that. Most of those measurements require instrumentation that the telescope just doesn't have.
We will identify interesting things to look at, and then these other instruments will go and look at them.
Q. Describe your relationship with NCSA.
A. My role here is to keep the connection between the science sidethe physics and the astronomy departmentsand the data management sidethe NCSA part, to keep that conversation going.
The physics department [at Illinois] sees cosmology as a field it wants to become more active in than in the past. I think the same is true in the astronomy department. These big cosmology projects [like LSST and the Dark Energy Survey] are a much bigger part of astronomy research than they have been.
Q. Why is NCSA valuable in that?
A. NCSA has capabilities, whichI wouldn't say uniquebut they're certainly rare. There aren't very many places that can manage datsets of the size that we're talking about and have the computer expertise of the kind we're talking about. And even some of the more general things that you'd think of as astronomy like visualization.
Just the management and visualization of these large datasets, NCSA is developing that expertise not just for astronomy but also for biology and other science applications. So NCSA plays an important role because a lot of the data-processing technology is applicable to a broad, diverse range of scientific applications, of which astronomy is just one.
Q. And those rare capabilities at NCSA that you describe are helping the astronomy and physics departments move into this particular niche?
The fact that NCSA's on the campus here means that we have an advantageous position. We can work on these projects. I can walk over to Don [Petravick] or Ray [Plante] at NCSA about various issues. It's true you can call people up on the phone, but it's better if you're 300 yards away instead of 3,000 miles away.
Back in the days of small science, and even nowadays in fields where the science is still small, you can do everything. You can propose the experiment, you can do the experiment, you can analyze the data, and it's all yours because it's a small project. Maybe it only cost $500,000, and you get a grant and you do it. When the project is $500 million, it's just different.
The upside is you get to do things you couldn't do otherwise.