Transcript: Irene Qualters from the NSF Discusses the NSCI Initiative

Print Friendly, PDF & Email
Irene Qualters, NSF

Irene Qualters, NSF

In this video from the 2015 HPC User Forum, Irene Qualters from the National Science Foundation discusses National Strategic Computing Initiative (NSCI). Established by an Executive Order by President Obama, NSCI has a mission to ensure the United States continues leading high performance computing over the coming decades. As part of the effort, NSCI will foster the deployment of exascale supercomputers to take on the nation’s Grand Challenges.”

Transcript:

“Good morning everyone, and thanks for the HPC user folks for inviting me. It’s really a pleasure to be back, and I haven’t seen many of you in quite awhile, but as I look across the audience I probably know 25 to 50% of the people here, so anyway it’s a pleasure to be here. Randy gave a great introduction and a perspective on the initiative, and I’m going to focus a little bit more– I’ll skip the overview slides. I wasn’t sure exactly where he was going to start, and I’ll just give you a more narrow view on NSF and how it sees it’s role and where it’s focus is.

Randy covered this. He mentioned that there are three lead agencies and with DOD and DOE, NSF’s role particularly is called out for scientific discovery advances, the broader HPC ecosystem and for work force development. I’m going to talk a little bit about what’s motivating us in those dimensions and I will readily admit this is very much early stage so it’s good to have this presentation now and I’m very interested in this group’s discussion and questions about how we might approach some of the objectives that I’m not touching on.

Again, Randy went through this and I’ll not spend time– Randy also talked about the objectives and I wanted to focus on objectives two, three and four. The press releases all focused on objective one, but for NSF the resonance for two, three, and four – while we’re interested in all the objectives – the resonance for two, three, and four is particularly strong. So I want to talk a little bit about that and I also wanted to say that for number five, we are at very early stages in nascent. So I really am interested in discussions from the groups here on how best we – as many agencies and particularly as NSF – should approach it.
So, also, a little bit on Next Steps, what’s going on now, there is an executive council that was mandated as part of the executive order with memberships from all the participating agencies. It’s co-chaired by OSTP and ONB. And we’re in the first phase of an implementation plan that was also mandated and the clock started July 29th. So 90 days later, we have to have our first implementation plan, so we’re working on that now.

I’m just going to go deep into number three. And say these are the kinds of activities, the areas within NSF, so I want to start and say first of all, this isn’t an IT initiative in NSF, this is an advancing science initiative. So the objective three, which is really a critical strategic issue for the nation, and is involving other agencies, and in particular IAARPA, and NIST are prominent. In the NSF world, engineering directorate, the computer information, and science and engineering directorate, the math and physical sciences directorate and the bio directorates are intimately involved, and I would say the first three right now are most active in looking at the post-war law.

As Randy indicated, this isn’t just hardware effort, it’s also rethinking the software activities. This isn’t an exhaustive list as I look here. Certainly the machine learning piece is really– should definitely be on this list, but this is an activity that I certainly see industry and academic engagement is– there’s already a baseline for that. Really trying to really lift that and really make it go faster and more profoundly is the goal. I think there’s a good basis for this one.

I want to talk a little bit– and I’m not going to get into this in great detail. But increasing that coherence – just stepping back a little bit and looking at it from the science perspective – the world of modeling and simulation, and I’ve used data science in place of data analytics and have given examples where science really cannot advance unless we make progress. And data simulation turns– that it’s in many disciplines is a rate-limiting activity.

Visualization, image analysis– there was just last week the president was in Alaska as part of the US chairmanship of the Arctic Council for the next two years, and one of the activities they talked about is doing a high resolution map of the entire Arctic, and that’s a huge image analysis program. They’ve just started I think the University of Illinois was kind enough to get them started by providing some time on blue waters to do this mapping, but the issue is that there are many images of the Arctic and many interests in the Arctic, but the images are all at different resolutions, and so trying to get a really first high resolution map is a non-trivial activity.

You could imagine that the systems we’re using to do that are not necessarily thought– were not necessarily designed or conceived to optimize for image analysis. But we can also see that in areas such as understanding the brain which is a very significant initiative in other areas of Geo, as well as looking at cosmology, this is a really critical function to be able to work well in order to advance science. So data compression as the large instruments, we need new algorithms, we need new approaches for compression. Certainly the visualization, particularly in the real-time arenas, also urgent. So trying to marry the capabilities of modeling and simulation often in a dynamic – not necessarily real-time but in a dynamic work flow with the data science attributes – is something that has profound science implications and will be an area of focus for us.

I also wanted to acknowledge that when I think of data science, and I’m going to credit Doug for suggesting a particular book that I used and we worked a little bit one of the images. This is a 2010 Drew Conway image of data science, and this resonates within the NSF community in that data science is emerging, and it’s inherently interdisciplinary. It’s pedagogical home is really unclear, but you need all the elements, you need the foundational underpinnings and statistics and mathematics. You need the domain knowledge to know which questions are relevant to ask and are important enough to ask, and you need the technical skills, the engineering practice to know what actually can be done.

Again, I think this plays into not only the second objective, but also learning and work force development in a very serious way. I’m going to credit Randy with this slide, just taking off what he mentioned in his presentation, the goal of this initiative is to move the state of the art, not just merge and combine, but really take the nation forward in both dimensions on a platform that has shared technology.

We also see a profound alliance between objective two and four which is increasing the capacity and capability, and these activities are– we have an internal working group that I’ll talk about. All directorates are participating in plans in this area. And there’s a good deal of excitement, and I want to again reference that for us, this is not just about technology or systems. The people on the software – in the software in particular – we see this as a critical opportunity to reinvent the software architecture, and the software base for the future.

So I’m a little bit, running out of time. I guess only one thing I’ll say here. The context of science is changing, and it’s not just science, everybody’s interaction with technology is profoundly changing the way everyone works, and even how our lives. Science is no different. I think examining this activity in that bigger context is really, I think, profound and strategic. And I think we can look at it from the science side. I think the role of industry – and how the broader industry and how the nation’s competitiveness can be enhanced – I think we’re really looking for a way to start that dialogue.

I think I’ll just close so I don’t take too much time. Right now our implementation is more on that second-wide bullet. This activity is being governed at the highest level within NSF, the heads of directorates. We have a cross-directorate working group that’s working on the initial planning. It’s being jointly led by size, and the math and physical sciences directorate. We also are establishing vehicles for community input. And those will certainly be through our advisory committees, which are public groups. And it will probably be through multiple advisory committees. Certainly the cyber-infrastructure one, but we expect other directorate advisory committees. And we also have a National Academy study that was looking at the future for advanced computing infrastructure. That final report is due in October, so that’s extremely timely. So this is really, for us, coming together in a way that it’s really set up to have high impact. And so this is the beginning, and we hope that you’ll engage and let us know how best we should include industry and should include the academic community.

I also want to acknowledge that the Council of Competitiveness has been looking at software and workforce development, and I think some of the work that that group has done will most certainly inform the approaches we take. But we’re looking to evolve it.”

Other Panelists:

Watch the complete Panel Discussion on NSCIRead the NSCI Fact SheetSee more talks from the HPC User Forum 

Sign up for our insideHPC Newsletter