In this video from the 2015 HPC User Forum, Will Koella from the Department of Defense 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:
“I want to kind of set the record straight a little bit. I’m actually coming in a little bit late to the process. I wasn’t involved with the creation of the NSCI or the staffing that went along with it. But certainly, where I was coming from helped track it.
Moving forward and working with Irene and our partners as a lead agency representative – working with Irene, our partners at DOE and OSTP, etc – to develop that implementation plan. That 90-day clock started on the 29th of July. Our first draft is due – what? Ten days, Irene? We’re cranking forward pretty heavily on that.
A little bit of background on me. I’m actually from NSA. We are the executive agent for DOD in its capacity as the lead agency for the NSCI. We can talk a little bit about why NSA versus some of the other aspects. Certainly, DOD has a large role across the board both within our agency, but also with the HPC Mod Program that you’re familiar with. I think the best description for us as why NSA has got a role in this NSCI should come as no surprise. We have significant investment in the HPC field, and across all objectives, really we have some degree at least a modicum of interest in making sure that they’re successful.
I guess the best description – and I’ll go a little bit more into detail on the objectives – but the best description of our role I think is going to be one of an advocate for the requirements process. When you look across the board, we’re ensuring that the National Security community in particular, but in general, those deployment agencies that don’t necessarily have the big budgets or the big deployment requirements are going to be incorporated. Whether you’re talking about law enforcement – obviously, the intelligence community, the military community – those requirements will be incorporated particularly on the data analytics side. DODs get a large mission requirements in that space.
We also are the lead development and deployment agency for objective three, obviously, once that starts coming online. Right now, that’s more of– we’re starting much further behind compared to the exascale parts, in the sense that we are really in the assessment phase of where we should even be investing in the first place. Investment across the board is very defused. There’s a lot of promise, but there’s also a lot of risk. I think one of the common characteristics is that there’s still a lot of miracles of science that need to occur for every one of those particular pathways before we could even contemplate a mission capable devices in any of those systems. But we’ve got a long history of being on the front end of that, so it made natural sense for us being a sense of that Nexus of national security. Bleeding edge research and development in terms of HPC, unique systems, etc. So that’s really where we fit in. And to be fair we also– I’m working across DOD, making sure that the requirements parts for HPC mod, certainly on the mod sim side, they’ve got a lot of interest in the exascale piece. And will be working with DOE to make sure that those requirements are put in. But also, I think this is starting to get into the implementation side, the role of the lead agencies.
A lot of the NSCI at the implementation level is going to be about essentially investment portfolio management. So we’re looking across the board of “Can we get more than one spin on from the same dime across the community?” which is traditionally with the federal government, we are notorious for not talking to each other. It should come at not surprise. So we’re trying to make sure that we can actually align our investments and for us, what that means particularly for NSA, what we are looking to kind of set as the model is we’re going to open up our kimono to DOE and to the other – NSF and folks like that – and say, “Here’s where were working, these are the folks, these are the areas of focus that we have. And what are you investing in, what are we investing in, can we align and synergize those investments across the board?” That’s, I think, the early faces of the NSCI implementation plan is going to be focused on that. What are we doing? Doing an assessment across the board? Or where are we spending out money now? Because FY16 is going to be more focused certainly on– It’s already set. We’re starting in 20 days. The budgets are set. We’re going to try to focus as best we can within our own agency budgets. And start aligning with the goal of influencing FY17 budget as best we can, but really focusing on FY18 for the year where we’ll start getting a big influx in terms of addressing gaps. Again, because we’re on that two year budget cycle, FY18 really that submission is going to be next spring, so we don’t have a lot of time to look across the board especially for objective three.
So we’re going to be very busy, I think part of Irene and I, we’re digging out our cat-herding chaps. We’re going to get that going. If you’ve every dealt in the inner agency– a little bit of background on myself, like Rob mentioned– I actually came from the hill. I was a professional staffer on the intelligence committee before I went up to NSA. So I’ve had plenty of experience from the inside of seeing how the sausage is made, and just how bad we can be. And we’re trying to avoid that. I think, just as a commentary on as more of a late comer to the NSCI, we feel it’s very important, obviously, I think we share your enthusiasm for it. And I have to say, it’s just one of the more elegantly put together and well-founded initiatives that I’ve seen coming out of the government. There was a lot of work that didn’t get necessarily put with the seven pages of the executive order, that really lay the foundation for the economic advantage, the security advantages, the scientific discovery that we are really excited about.
Going into a little bit of detail as to where we think our roles are going to fit within the particular objectives. Obviously we are going to support DOE’s effort, they’re taking the lead on objective one and delivering that exascale system. We are going to make sure that the HPC mod programs and the rest of the community, especially the mod sim and the scientific community can get those requirements in and where we’re making investments, actually help accelerate the program. That would be a goal I think you can see. To do a little bit of a divide and conquer, which is going to make Doug’s project management job a little bit more complicated but we’re going to try to help where we can.
Objective two is one where we are– certainly an area where we are going to take a balance role in between us, NSF and DOE, bringing up the data analytics side, as you can imagine, that’s an area of huge importance for us right now. We have always traditionally been in a more specialized end of high-performance computing but HPDA is a growing area for us and in particular getting that, making sure that the data analytics side for IC problems, for NSA’s problems, for the military and law enforcement community. There’s tons of applications out there that we need to make sure is getting incorporated, but we also have ecosystem issues. How do we– for us, our high-performance computing missions are increasingly taking place in the more open intelligence analyst production environment and so those environments have got to be more tightly integrated. It’s been nice to see some of the discussions that have gone on in the last couple of days where folks are seeing that issue as well. So that’s about the core infrastructure, it’s about the software. Greater degrees of abstraction working there. We see that focus and where we’re going to help along.
The other piece I have a talk a little bit about our specific role going in. Irene mentioned that the executive counsel meeting met last on the 26th of August. And a couple things that they laid out was in recognition, certainly an objective three, which is where I think we’re taking a more of a leadership role in the deployment and development side. Obviously, there’s a lot that’s going to have to come out of that. But recognizing that there was going to have to be a lot more effort in trying to identify areas of potential investment for post CMOS, post Moore’s Law. They established a future computing technology as executive counsel underneath the overarching executive counsel. That is going to be focused solely on and picking out where we are be investing across these diffused technology paths with a goal of trying to identify a candidate path, or paths that we’re going to look at. Try to identify something in the next three to five years that would give us a viable alternative to CMOS. Looking again, 10, 15 years down the path.
For us it’s really important– I think traditionally that the long lead time for some of the technologies that we’re looking at, we’re taking a parallel approach in terms of trying to get the most out of Moore’s Law. So the initial focus is going to be on technologies that will extend the classical, digital computing architecture, and framework for as long as we can into the future, because we hope we can get 10 years out of it. While simultaneously trying to identify alternative paths that will help us keep going on that path. And I think realistically what you’ll see is that it’s not necessarily going to be a one-to-one replacement. I don’t think we’ll ever find something. Certainly right now we don’t see anything that will replace classical, digital computing for a lot of applications, but there may be some specialized places, whether it’s neuromorphic, whether it’s chemical-biological based computing, or quantum computing where we can take it in a different direction and continue that basis.
And finally something like that that comes about, that is such – like Randy mentioned – because those fields are certainly in the pre-competitive stage for you guys it doesn’t make monetary sense, necessarily, to invest in those spaces. You need the national security or the federal government imperative to derive the investment. But one of the underlying, most important aspects of the initiative is that we want to be the lead in developing “the next thing” for 15 years down the road. So those new architectures we want to make sure – even though they are being developed for the government – we have got to plan to transition it. The objectives four and five, which gets a little bit out of sight of NSA’s comfort zone in terms of dealing with the public. We don’t like to talk, right? But they sent me because I can look at your shoes instead of being the– we’re notoriously introverted. But it’s hugely important that we work with everyone to facilitate that transition ultimately when it does occur. So all of those points being, we’re supporting the FCT Council. I’ve actually stood up a joint program office between the lead agencies, myself – excuse me – myself, DOE, NSF are going to be kind of the executive coordinators across the inter-agency for the future computing aspects as we move forward.
And if you read the executive order, one of the things that wasn’t necessarily– it was not explicitly noted in the EO but certainly is important in some of the supporting documentation is, national security is clearly an application space where we need help. So we’ve got a clear leadership role there because we bridge the gap between DOD and the IC and the data analytics team, and we’re kind of there. So as that goes along, we’re kind of the Johnny-on-the-spot filling in the finger in the dyke across the board, so between us– I think Irene and I are probably going to be the busiest – no offense to the DOE guys, they’ve got the benefit of some focus – we’re going to be jumping across the boards I think, but looking forward to the challenge.
With that I think we can mention key partners. As part of the initiative, there’s certainly, across the board, other folks that are going to be taking part in it. We’re working with IR, we mentioned NEST looking at that future computing space finding out the technologies that are going to be the most important to invest in. There’s a lot of good work that’s going on there, and we’re looking forward to where this is going to go. I think some of the language that you used, or one of the articles that we saw was the hopes that this will be an Apollo project for computing for the nation. We certainly– I hesitate to say that we will put a man on the moon. We would like to see that. It’s hard to go– we recognize that it’s hard to grasp the American public’s imagination with computers, but I think what we’re going to look to you guys to help us with is capturing that challenge in tangible ways that what’s the impact of where this is going to go. I was talking to Joseph yesterday, I think we had some great ideas where we want to be able to cure cancer. We want to put a man on Mars. We want to do these things. You’re not going to be able to do that without a super computer and continue to push the path forward.
Finally, I would reiterate my last note before I go, and I already mentioned this a little bit, but we want to put the challenge to you guys particularly for objective five. As a former congressional staffer, I’ll say I’m skeptical of the federal government coming down from the mountain with the Ten Commandments about how the best way for the public sector to contribute. What I really feel like there’s an implicit leadership role for the community, for industry and academia, to help us find those ways. We’re looking for you to tell us how we can help you whether it’s incentivization, direct investments, etc. That’s our challenge to you going forward, and if you have any questions come up, that’s all I’ve got.
“
Other Panelists:
- Randy Bryant, OSTP * Video/Transcript
- Irene Qualters, National Science Foundation * Video/Transcript
- Doug Kothe, Oak Ridge National Laboratory * Video/Transcript
Watch the complete Panel Discussion on NSCI * Read the NSCI Fact Sheet * See more talks from the HPC User Forum