Transcript: Doug Kothe from Oak Ridge Discusses the NSCI Initiative

Print Friendly, PDF & Email
Doug Kothe, ORNL

Doug Kothe, ORNL

In this video from the 2015 HPC User Forum, Doug Kothe from Oak Ridge 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:

“What I’m showing now is just a few slides that Steve Binkley showed at his advisory committee meeting in July. And so– it’s known as ASCAC, for the ASCR Advisory Committee. You can find this presentation online. I’m just going to show a few slides from this presentation. Also nice presentation from Dan Reed, a subcommittee of ASCAC that gave some very good recommendations to the Department of Energy in this case the ASCR program office on how to proceed with the exascale computing initiative.

The DOE role is being embodied within one project known as the exascale computing initiative or ECI. This is a project with a capital P, and I’ll explain in a little bit. I’ll also point out that the exascale research and funding in the DOE has been ongoing for many years in fact. This is essentially not a new start but it is in that it’s being projectized. In DOE land that means we’re following a fairly rigorous process known as four thirteen. Historically that’s a project process for large capital acquisitions. This particular initiative ECI, as we’re calling it, is a ten-year project formally starting October 1. Okay, and it does have as part of it, capital acquisitions, namely, two or more capable exascale systems. In fact I’ll back up a little bit with regard to NSCI. DOE, at least the way I see it – this is Doug speaking now – can and will contribute to all the NSCI objectives.

The one that it is most on point for is delivering the capable exascale systems. And in my role – which I’ll get into maybe in the next panel a little bit more – is more on the application side, and I’m using the term application for large simulation codes. The way I see it, DOE needs to and will deliver break-through science, energy, and national security applications that are capable of exploiting these exascale systems to deliver new insights and solutions.

So I’ll now go to this slide, and I think I just have one more. Things have changed a little bit, but the ECI is DOE’s implementation of the NSCI, so the ECI is a project that will be executed by the DOE lab personnel. You see in Steve’s slide here there are 17 DOE labs, and all will play a role. These six are most likely to play leadership roles, although from what I know of my colleagues across the DOE lab complex, there are lots of very talented people who have been deeply embedded into HPC, and we’re essentially looking for– essentially right now we’re assembling a leadership team for the ECI. You see the six labs that are listed, three NSA labs and Oak Ridge, Argonne, in actually Berkeley have been more involved recently with Steve and Thuc in formulating the initial plans.

I am here in part because the project office will reside at Oak Ridge, and in that sense Oak Ridge will carry a lead responsibility and a lead role in executing the project. That does not mean that all Oak Ridge personnel will lead this project. It is a very complicated project, and we need essentially the best and brightest across the DOE lab system, but we are in the process of establishing a project office and putting together a leadership team, and I say we as the DOE. I’ve been engaged with this group for about four months now, and in terms of a 04/13 project process, it’s onerous. There are lots of good reasons to do this.

We’re viewing this as a very complicated multi-institutional, multi-disciplinary project. Because of that, we’re going to follow a fair amount of rigor. In this case we’re going to go through– if you’re familiar with this progress, great! Maybe you can enlighten me a little bit more. But, we’re going to go through several critical decisions, but right now, we’re in the process of what’s called CD-0, which is to get formal mission-need approved and this comes with some important documents and presentations and critical reviews by federal personnel, and so right now, we’re in the process of seeking CD-0 or mission-need approval. The next step would be CD-1, Critical Decision 1, which essentially gets into the acquisition of strategy, the alternatives-analysis and essentially having what’s equivalent to a baseline project plan.

You can see in Steve’s slide there that we DOE lab personnel, working with the ASCR program office and the ASC office have put together essentially a project structure, and I think if you look at the structure, you can get a sense of the scope of the project and what we’re endeavoring to undertake with regard to the kind of the breadth and depth of the RND.

My role currently is lead for the applications development portion of the ECI. Okay, so you see the work break down structure and having a lot of experience with large projects and programs I can tell you that this structure will likely change. It’ll likely change as we learn more and we begin to execute.

Let me just point out the four main technical areas. Application development – which is my role – is going to emit, we believe, dozens of applications that’ll be exascale ready. And the traditional – what I’ll call – the traditional applications are certainly in scope. If you ask, “Is there one killer app?” I would say “No”, there’s probably dozens from what we can see. Historically, you’re three killer apps – at least as I would see it – would be those in support of the stock pile, would be those in support of crypto analytics and being a nuclear engineer, I would argue that nuclear reactors have been a killer app since the 40’s. They have been, will be, are, and will be, but we see dozens of other applications, and I think I’ll reserve the next panel session to dive into the application space in a little bit more detail, but the interesting thing about this project on the application side, we’re looking at the applied energy offices in the Department of Energy as well as the Office of Science, and also reaching out to potentially other federal agencies in terms of a shared fate formal engagement.

In particular– I think it was noted by Randy and Irene, obviously there are a lot of synergies and reasons to collaborate with NSF, but also with NIH, and potentially Noah and NASA.

I think at this time, I’m not going to get into the details of these elements until and unless you have questions. Let me just say a little bit more about applications and then conclude. On the application side, a RFI was submitted by the Department of Energy to all the DOE labs in early June, for white papers from DOE Lab personnel with regard to their ideas and thinking’s about what applications can be, should be developed, what’s their decadal problem, what’s the impact urgency, what are the challenges. Those white papers, we received about 130 of them from 12 of the 17 DOE labs, and they’re still coming in. Another RFI, I believe is due to go out this week to other agencies, in particular NSF and NIH. So we’re currently gathering information to help formulate and influence our plans and our scope. So, on the application side, I do encourage you to – if not contact me, you can contact one of the program offices. It’s never too late to get your good ideas with regard to the exascale application side.

And, again, there’s a software technology piece. And you can see all the elements of the software technology piece. Hardware technology piece that really is carrying on the fast forward and design forward sorts of activities with the vendors. So there’s a lot of scope and resources available for those activities and hardware technology. And exascale systems will essentially deploy two or more diverse systems within the next ten years. Nominally, 2023, but don’t hold me to that. We’ll have to see what happens. But this project is expected to run through 2025. With that said, I’ll conclude for now and be happy to answer your questions later. Thank you.”

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