NASA’s Jet Propulsion Laboratory is moving into the cloud.
Today we bring you the first segment of our three-part chat with Tom Soderstrom, the IT CTO at NASA JPL and Khawaja Shams, senior solution architect at NASA JPL.
We start off by asking the obvious — why cloud, and why now?
TS: What we did is, about two years ago, we looked at the industry trends that were coming up, and we came up with seven — and cloud computing was one of them. JPL does early exploration of technologies to see if they can benefit the scientists and technologists. So, we started looking at it two years ago and it became very relevant, because we have a tremendous increase in demand in IT for the mission.
We have the busiest period ever coming up here for JPL within the next two years, so we need more computing, more storage, more data centers, and much more flexibility and speed. We want to run and store data and programs wherever it’s most appropriate, and the cloud looks like it would have that potential for us. So, that’s why we started looking at it.
KS: That was basically our reasoning, and then we also realized that a lot of our missions and scientists had very elastic needs for computing . . . where they would need a lot of computation for a few hours in a day, and then their demand would shrink down to basically a very few number of computers for the rest of the day. We found that the cloud can offer very economical options to leverage elasticity.
FCB: You are out in California. We’re here in D.C. Have you gotten any specific guidance or have you shared best practices with NASA [out here]? We recently spoke with NASA CIO Linda Cureton, and she just talked generally about what she’s doing at NASA here in D.C. When it comes to cloud computing. Have you shared best practices . . . or, if not, do you plan on doing that in the future?
TS: Yes, indeed. We are sharing. In fact, our whole approach is about being proactive and exploring multiple clouds and share our lessons learned — the good and the bad — so we can advance the state of the practice, but also so we can learn and not make big mistakes. Linda Cureton is a believer in cloud, and so is my boss Jim Rinaldi, CIO of JPL. We share best practices with all the NASA centers, but also with other industry players, including internationally. We’re working with ESA, the European Space Agency, to try to see how we could collaborate better. So, we share best practices and we share our vision and our approach, as well as the results and road map. That way we can come to a better cloud readiness level.
FCB: Overall, what are the benefits of cloud computing for JPL. The only reason we ask is because we’ve talked to other agencies . . . and some are a little bit wary so far. . . . How did you make the decision with cloud and what are the benefits?
TS: That’s really the crux of the matter for us. . . . Two years ago we looked at the cloud and saw it had all of these potential benefits, and we worked a lot with Computer Science Corporation (CSC) and we studied the cloud. So, we saw, in depth, what the potential benefits were and the potential drawbacks were, but every institution is different. What we wanted to do was to keep it real for our scientists and technologists. We had to get in and explore early.
We are on the other side of the coin, I would say . . . [because] we believe in proactive, early exploration so we can understand what the net benefits are for JPL and our personnel, as well as for NASA. We can’t do that unless we’re in there trying it. That was our overall approach — to just get started and to try it. . . . Two years later, I think we have a pretty good handle on what the real net benefits will be and are, and what the real drawbacks are. We are seeing some real benefits.
The big issue was security. Could we actually use the cloud right away in a secure way?
What we did was come up with something we call the ‘wheel of Security’. We divided our data — or tagged it — and asked, is it public data, secure data and is this something we can do today? When we did that, we discovered that we have a lot of public data and we can do our processing [in the cloud] both to benefit from it immediately, but also to learn.
Now what we’re doing is looking more at the private data, if you will and this wheel of security — we’ll talk to the vendors and providers — we’ll spin it and when they are ready to take [it] on, we’ll [move] the next big chunk.
KS: We had to process 184,000 images of Saturn for a public outreach project, and we started processing it in our labs and after a couple of weeks we decided it was going to take too long. We were only using one machine to run it. We put it out on the cloud and we were able to process the same images in about 5 hours.
So, the dramatic increase of 15 days to 5 hours is very worthwhile as a strategic research advantage. One of the things that folks have pointed out is, if you have a computation that requires 100 hours of computation, if you run it on one machine for 100 hours or 100 machines for one hour, it literally costs you the same thing, except you gain 99 hours of more research time. We want to bring this benefit down to our scientists.
We have already some of these successes come through with our public data processing.