WEBVTT 1 00:00:00.240 --> 00:00:03.770 A:middle L:90% deep in our department. Uh and she is currently 2 00:00:03.770 --> 00:00:06.780 A:middle L:90% the program manager for the A. S. C 3 00:00:06.790 --> 00:00:10.490 A:middle L:90% . R. Program the advanced Scientific Computing research program 4 00:00:10.500 --> 00:00:13.990 A:middle L:90% in the Department of Energy's Office of Science. Let 5 00:00:13.990 --> 00:00:16.269 A:middle L:90% me tell you a little bit about Dr Nowell's career 6 00:00:16.280 --> 00:00:19.929 A:middle L:90% because she has had an interesting career. Her dissertation work was 7 00:00:19.940 --> 00:00:24.410 A:middle L:90% done here in the digital libraries research lab and her 8 00:00:24.410 --> 00:00:27.070 A:middle L:90% work involved the design of one of the first information 9 00:00:27.079 --> 00:00:31.649 A:middle L:90% visualization user interfaces for the envision project. This was 10 00:00:31.649 --> 00:00:37.439 A:middle L:90% a joint project funded by NSF sponsored by them to 11 00:00:37.619 --> 00:00:40.250 A:middle L:90% which was one of the early digital library projects for 12 00:00:40.250 --> 00:00:44.600 A:middle L:90% CS literature. Most of you have used the MTL 13 00:00:44.609 --> 00:00:47.899 A:middle L:90% and the Tripoli Computer Society deals and this was a 14 00:00:47.909 --> 00:00:52.140 A:middle L:90% precursor project. Um This all happened in the mid 15 00:00:52.140 --> 00:00:55.740 A:middle L:90% 90s and the goal of this project was to explore 16 00:00:55.750 --> 00:00:59.640 A:middle L:90% innovative ways of searching and viewing information. So if 17 00:00:59.640 --> 00:01:03.850 A:middle L:90% you think about it, that's A little bit more 18 00:01:03.850 --> 00:01:06.629 A:middle L:90% than 15 years ago. Think about how we take 19 00:01:06.640 --> 00:01:10.019 A:middle L:90% for granted the deals that we use today. Dr 20 00:01:10.019 --> 00:01:11.510 A:middle L:90% Noel. Previous to that served as a chief scientist 21 00:01:11.510 --> 00:01:15.260 A:middle L:90% in the information analytics group at the Pacific Northwest National 22 00:01:15.260 --> 00:01:19.159 A:middle L:90% Lab. Some of us noticed by the acronym P 23 00:01:19.159 --> 00:01:19.859 A:middle L:90% . N. N. L. And then she 24 00:01:19.859 --> 00:01:23.420 A:middle L:90% moved to be a program director in data data analysis 25 00:01:23.430 --> 00:01:27.150 A:middle L:90% visualization in the office of cyber infrastructure at NSS. 26 00:01:27.640 --> 00:01:32.159 A:middle L:90% Before she joined them mentioned, she also has had 27 00:01:32.159 --> 00:01:34.810 A:middle L:90% experience as a faculty member. Early in her career 28 00:01:34.810 --> 00:01:37.219 A:middle L:90% she was a faculty member at Lynchburg College where she 29 00:01:37.230 --> 00:01:41.030 A:middle L:90% headed both the theater department and then subsequently was Sheriff 30 00:01:41.030 --> 00:01:45.730 A:middle L:90% Computer. So probably unique in our country, Held 31 00:01:45.730 --> 00:01:49.709 A:middle L:90% the chairmanship of those two are highly accomplished. The 32 00:01:49.709 --> 00:01:55.329 A:middle L:90% mother clearly has interdisciplinary talents and interests and we bought 33 00:01:55.329 --> 00:01:59.930 A:middle L:90% our career service in to the scientific community serving in 34 00:01:59.939 --> 00:02:04.439 A:middle L:90% various positions in government institutions. We will be presenting 35 00:02:04.439 --> 00:02:06.760 A:middle L:90% her as a department with a C. S. 36 00:02:06.760 --> 00:02:08.490 A:middle L:90% F. B. T. Distinguished Alumni Award later 37 00:02:08.490 --> 00:02:10.530 A:middle L:90% today. I'm very, very proud and happy to 38 00:02:10.539 --> 00:02:14.050 A:middle L:90% have you with us. Please help me welcome dr 39 00:02:14.229 --> 00:02:15.699 A:middle L:90% back to C. S. F. P. 40 00:02:15.699 --> 00:02:23.750 A:middle L:90% T. Thank you so much for the warm welcome 41 00:02:23.439 --> 00:02:27.349 A:middle L:90% . Those kind words, those of you in the 42 00:02:27.360 --> 00:02:30.270 A:middle L:90% back please let me know if at any point you 43 00:02:30.270 --> 00:02:31.909 A:middle L:90% can't hear, give me some kind of signal that 44 00:02:31.909 --> 00:02:36.000 A:middle L:90% I need to be louder. One advantage to the 45 00:02:36.009 --> 00:02:38.330 A:middle L:90% theatrical background as I can probably adjust and fill the 46 00:02:38.330 --> 00:02:42.439 A:middle L:90% ballroom if I need to, but I'm trying not 47 00:02:42.439 --> 00:02:44.659 A:middle L:90% to deafen people up front if I don't have to 48 00:02:46.939 --> 00:02:49.620 A:middle L:90% when I was here, no one would have ever 49 00:02:49.620 --> 00:02:53.360 A:middle L:90% guessed. Certainly not me that I would wind up 50 00:02:53.840 --> 00:02:58.800 A:middle L:90% Running programmes in an office of supercomputing or two or 51 00:02:58.800 --> 00:03:00.699 A:middle L:90% do many of the things that I have done and 52 00:03:00.699 --> 00:03:02.900 A:middle L:90% I want to start with a thank you to the 53 00:03:02.909 --> 00:03:07.430 A:middle L:90% faculty members here who taught me so well and prepared 54 00:03:07.430 --> 00:03:12.060 A:middle L:90% me to go places that I never expected to go 55 00:03:12.840 --> 00:03:15.000 A:middle L:90% and to be doing. The things that I'm doing 56 00:03:15.000 --> 00:03:17.449 A:middle L:90% now, that we're not within my wildest dreams. 57 00:03:17.449 --> 00:03:20.659 A:middle L:90% When I was a student here at Virginia Tech, 58 00:03:23.740 --> 00:03:25.229 A:middle L:90% I'm going to talk a little bit about the Office 59 00:03:25.229 --> 00:03:30.310 A:middle L:90% of Science, the Office of Advantage Scientific Computing research 60 00:03:30.310 --> 00:03:34.960 A:middle L:90% that we usually call Oscar. I'm going to ground 61 00:03:34.960 --> 00:03:38.340 A:middle L:90% some of the discussion about data in a discussion of 62 00:03:38.349 --> 00:03:40.599 A:middle L:90% climate change and explain to you why that matters to 63 00:03:40.599 --> 00:03:45.050 A:middle L:90% the Office of Science. And I hope on a 64 00:03:45.050 --> 00:03:46.909 A:middle L:90% more personal level, why it matters to me and 65 00:03:46.909 --> 00:03:50.460 A:middle L:90% should matter to every human being on the planet. 66 00:03:52.139 --> 00:03:53.500 A:middle L:90% And then I'll talk a little bit about the research 67 00:03:53.500 --> 00:03:58.000 A:middle L:90% that lies ahead and the work that I hope you 68 00:03:58.000 --> 00:04:02.500 A:middle L:90% will do with us. I say this and having 69 00:04:02.500 --> 00:04:06.360 A:middle L:90% conversed with Barbara about what to present here, I 70 00:04:06.360 --> 00:04:10.169 A:middle L:90% no longer do my own research. It's probably a 71 00:04:10.180 --> 00:04:14.819 A:middle L:90% decade since I was a hands on researcher in visualization 72 00:04:14.819 --> 00:04:18.779 A:middle L:90% and the cognitive issues associated with information visualization and scientific 73 00:04:18.779 --> 00:04:23.569 A:middle L:90% visualization. My work these days is not in designing 74 00:04:23.569 --> 00:04:27.730 A:middle L:90% interfaces, but in designing and leading large programs of 75 00:04:27.730 --> 00:04:30.629 A:middle L:90% research for the country. And now, particularly for 76 00:04:30.629 --> 00:04:34.180 A:middle L:90% the Office of Science, the Office of Science is 77 00:04:34.180 --> 00:04:39.480 A:middle L:90% not well known to many faculty and graduate students, 78 00:04:39.939 --> 00:04:43.480 A:middle L:90% but we are one of the nation's largest funders of 79 00:04:43.480 --> 00:04:46.649 A:middle L:90% research in the physical sciences. We operate a number 80 00:04:46.649 --> 00:04:50.560 A:middle L:90% of scientific user facilities, 31 to be exact across 81 00:04:50.560 --> 00:04:54.319 A:middle L:90% all of the Office of Science and the users of 82 00:04:54.319 --> 00:04:58.329 A:middle L:90% those facilities fan not only this country, but other 83 00:04:58.329 --> 00:05:04.259 A:middle L:90% continents. We support people in universities and industry in 84 00:05:04.259 --> 00:05:08.750 A:middle L:90% wide ranges of research and national labs and using the 85 00:05:08.750 --> 00:05:14.449 A:middle L:90% machines. Over 27 0 researchers who have doctorates or 86 00:05:14.449 --> 00:05:17.029 A:middle L:90% are graduate students and postdocs have sort of support of 87 00:05:17.029 --> 00:05:23.579 A:middle L:90% one kind from the Office of Science. Within the 88 00:05:23.579 --> 00:05:27.420 A:middle L:90% Office of Science. The goals my office are to 89 00:05:27.420 --> 00:05:31.459 A:middle L:90% support leadership and computational science and super cute computing, 90 00:05:32.040 --> 00:05:35.360 A:middle L:90% both through operation of machines and through development of the 91 00:05:35.360 --> 00:05:42.790 A:middle L:90% software that runs on those machines. The office operates 92 00:05:42.800 --> 00:05:47.050 A:middle L:90% to leadership class supercomputing facilities at Oak Ridge National Lab 93 00:05:47.060 --> 00:05:50.910 A:middle L:90% , Argonne National Lab. Those are among the fastest 94 00:05:50.910 --> 00:05:55.160 A:middle L:90% machines in the world. And we operate a very 95 00:05:55.170 --> 00:06:00.839 A:middle L:90% practical capacity machine at the National Energy Research Supercomputing Center 96 00:06:00.839 --> 00:06:04.959 A:middle L:90% Nurse along with Energy Sciences Network S. Net, 97 00:06:05.339 --> 00:06:09.550 A:middle L:90% which connects all of those machines to our national labs 98 00:06:09.550 --> 00:06:12.160 A:middle L:90% and occasionally in university or two. Along the way 99 00:06:14.639 --> 00:06:17.629 A:middle L:90% the computer science program, he is focused on two 100 00:06:17.629 --> 00:06:23.019 A:middle L:90% things Right now, we focused inward at providing the 101 00:06:23.019 --> 00:06:27.560 A:middle L:90% operating systems, runtime systems and programming environments and tools 102 00:06:27.939 --> 00:06:31.439 A:middle L:90% for the supercomputers that we operate. And then we 103 00:06:31.439 --> 00:06:35.399 A:middle L:90% focus on the science requirements of the other offices within 104 00:06:35.399 --> 00:06:39.800 A:middle L:90% the Office of Science, helping them make use of 105 00:06:39.810 --> 00:06:44.670 A:middle L:90% those systems to carry out their computational science and analyze 106 00:06:44.670 --> 00:06:47.399 A:middle L:90% the data that comes in increasing volumes from the user 107 00:06:47.399 --> 00:06:51.629 A:middle L:90% facilities that they operate. This is within the context 108 00:06:51.629 --> 00:06:56.120 A:middle L:90% of an office that has a number of components. 109 00:06:56.129 --> 00:07:01.000 A:middle L:90% We have a networking and collaboration program at an applied 110 00:07:01.000 --> 00:07:05.040 A:middle L:90% mathematics program, both of which fund lines of research 111 00:07:05.040 --> 00:07:08.170 A:middle L:90% that look a lot to me like things I saw 112 00:07:08.170 --> 00:07:11.300 A:middle L:90% in the computer science department here that we put them 113 00:07:11.300 --> 00:07:14.459 A:middle L:90% in separate programs and that's probably better for their budgets 114 00:07:15.439 --> 00:07:17.680 A:middle L:90% . And we have a program called Side Ac that 115 00:07:17.680 --> 00:07:23.850 A:middle L:90% Scientific Discovery through advanced computing that operates in partnership with 116 00:07:23.850 --> 00:07:28.370 A:middle L:90% the other offices and enables domain scientists to get funding 117 00:07:28.379 --> 00:07:30.879 A:middle L:90% from other offices to work with the computer science and 118 00:07:30.879 --> 00:07:35.250 A:middle L:90% math folks that we support transitioning software that we develop 119 00:07:35.339 --> 00:07:42.040 A:middle L:90% into operational use for science. This is a summary 120 00:07:42.050 --> 00:07:46.790 A:middle L:90% of the budget fantasy That we had last summer when 121 00:07:46.790 --> 00:07:49.620 A:middle L:90% we had some illusions that our budget would go up 122 00:07:49.620 --> 00:07:54.870 A:middle L:90% about$15,000 in the coming year. Instead, we 123 00:07:54.870 --> 00:07:59.779 A:middle L:90% are Given a baseline of the 2012 budget which was 124 00:07:59.779 --> 00:08:03.819 A:middle L:90% around$440 million dollars from which thank you. Sequestration 125 00:08:03.819 --> 00:08:07.230 A:middle L:90% . We are down somewhat and expect to be down 126 00:08:07.230 --> 00:08:11.529 A:middle L:90% at least for the coming year. Those who don't 127 00:08:11.529 --> 00:08:15.800 A:middle L:90% know they sequestration is a 10 year sentence, Not 128 00:08:15.800 --> 00:08:18.290 A:middle L:90% a single one. So we continue to hope that 129 00:08:18.300 --> 00:08:22.100 A:middle L:90% flat will not remain the new up someday there will 130 00:08:22.100 --> 00:08:26.759 A:middle L:90% be budget increases again. But meanwhile we move on 131 00:08:30.540 --> 00:08:33.500 A:middle L:90% . Simulation modeling and simulation are critical to the mission 132 00:08:33.500 --> 00:08:37.230 A:middle L:90% of the Office of Science. If you think about 133 00:08:37.240 --> 00:08:41.289 A:middle L:90% energy and the things that relate to energy that will 134 00:08:41.299 --> 00:08:46.230 A:middle L:90% give you some insight into what we care about and 135 00:08:46.240 --> 00:08:50.570 A:middle L:90% why we care about ways of producing energy, things 136 00:08:50.570 --> 00:08:56.259 A:middle L:90% like combustion and nuclear power plants. Of necessity because 137 00:08:56.259 --> 00:09:01.240 A:middle L:90% of our partnership with the National Nuclear Security Administration. 138 00:09:01.250 --> 00:09:05.409 A:middle L:90% We care about maintenance and monitoring of the nation's nuclear 139 00:09:05.409 --> 00:09:09.539 A:middle L:90% weapons stockpile, which is managed largely and monitored largely 140 00:09:09.549 --> 00:09:15.330 A:middle L:90% through modeling and simulations on supercomputers. Today we care 141 00:09:15.330 --> 00:09:18.779 A:middle L:90% about climate change and the environment, not just for 142 00:09:18.779 --> 00:09:22.340 A:middle L:90% the personal and obvious social reasons, but because ways 143 00:09:22.340 --> 00:09:28.299 A:middle L:90% of producing and consuming our energy resources impact climate change 144 00:09:28.299 --> 00:09:31.259 A:middle L:90% tremendously and impact the environment we all live in in 145 00:09:31.259 --> 00:09:35.230 A:middle L:90% many other ways. But also because through manipulation of 146 00:09:35.230 --> 00:09:39.419 A:middle L:90% environmental environmental resources we may be able to achieve carbon 147 00:09:39.419 --> 00:09:45.330 A:middle L:90% sequestration and even mitigate some of the impact of nuclear 148 00:09:45.330 --> 00:09:50.080 A:middle L:90% waste. We need um a stream that's X. 149 00:09:50.080 --> 00:09:54.350 A:middle L:90% Escape because it will take three eggs a flops to 150 00:09:54.350 --> 00:09:56.649 A:middle L:90% completely model the U. S. Power grid. 151 00:09:58.139 --> 00:10:01.710 A:middle L:90% Also one of our major concerns. And it will 152 00:10:01.710 --> 00:10:03.509 A:middle L:90% take an extra scale machine to allow us to do 153 00:10:03.509 --> 00:10:07.649 A:middle L:90% climate modeling in a scale of roughly 10 m. 154 00:10:07.139 --> 00:10:11.789 A:middle L:90% Which is our ambition. When we look at our 155 00:10:11.789 --> 00:10:16.240 A:middle L:90% exa scale computing initiative, this is our path to 156 00:10:16.240 --> 00:10:18.799 A:middle L:90% the future. This will be a partnership across the 157 00:10:18.799 --> 00:10:22.960 A:middle L:90% government and our national labs, the computing industry, 158 00:10:24.340 --> 00:10:31.820 A:middle L:90% our labs. Everyone, if we were to build 159 00:10:31.830 --> 00:10:37.240 A:middle L:90% an extra scale system from today's components, It would 160 00:10:37.240 --> 00:10:41.059 A:middle L:90% consume over 200 MW of power which is completely unacceptable 161 00:10:41.059 --> 00:10:45.409 A:middle L:90% in terms of its hazards to the people in the 162 00:10:45.419 --> 00:10:48.309 A:middle L:90% building Who had operated. And also because that would 163 00:10:48.320 --> 00:10:50.909 A:middle L:90% cost us$200 million dollars a year. Which is 164 00:10:50.909 --> 00:10:56.240 A:middle L:90% more than the entire research budget for us. So 165 00:10:56.240 --> 00:10:58.730 A:middle L:90% we have an ambitious program of research to produce a 166 00:10:58.730 --> 00:11:05.789 A:middle L:90% machine That consumes only 20 MW of power and yet 167 00:11:05.799 --> 00:11:11.509 A:middle L:90% runs at Exxon flops And fits within 500 cabinets while 168 00:11:11.509 --> 00:11:16.759 A:middle L:90% also being programmable and understandable despite billion way concurrency. 169 00:11:18.440 --> 00:11:22.029 A:middle L:90% Most people don't live in a world where they think 170 00:11:22.029 --> 00:11:24.919 A:middle L:90% about exa flops a lot. One way to think 171 00:11:24.919 --> 00:11:28.929 A:middle L:90% about exit flops is on the order of a million 172 00:11:28.929 --> 00:11:33.929 A:middle L:90% flops per person on the planet. Think about how 173 00:11:33.929 --> 00:11:39.389 A:middle L:90% well we control the population of the planet and whether 174 00:11:39.389 --> 00:11:41.629 A:middle L:90% we really want processors at war with each other on 175 00:11:41.629 --> 00:11:46.120 A:middle L:90% a supercomputer or do we need new ways to conceptualize 176 00:11:46.129 --> 00:11:52.860 A:middle L:90% management our systems at that scale. But it all 177 00:11:52.860 --> 00:11:56.830 A:middle L:90% comes back to energy efficient computing Right now. one 178 00:11:56.830 --> 00:12:03.110 A:middle L:90% Petaflop uses three megawatts here,$3 million a year 179 00:12:03.120 --> 00:12:07.500 A:middle L:90% for the power. We are sometimes asked why why 180 00:12:07.500 --> 00:12:09.870 A:middle L:90% ? Instead of building an exa scale machine, don't 181 00:12:09.879 --> 00:12:13.610 A:middle L:90% we build 10 machines that are 100 peta flops. 182 00:12:15.440 --> 00:12:18.100 A:middle L:90% You don't solve the power problem right. You still 183 00:12:18.100 --> 00:12:22.659 A:middle L:90% have a power bill that's larger than our research budget 184 00:12:22.539 --> 00:12:26.210 A:middle L:90% . So we know that we have to change the 185 00:12:26.220 --> 00:12:31.240 A:middle L:90% way computing has done and that has tremendous implications across 186 00:12:31.250 --> 00:12:35.440 A:middle L:90% all of science. This is an old slide that 187 00:12:35.440 --> 00:12:39.409 A:middle L:90% was put together by our advisory committee. I want 188 00:12:39.409 --> 00:12:43.259 A:middle L:90% to pause on for a minute despite a couple of 189 00:12:43.269 --> 00:12:46.320 A:middle L:90% errors in it because it really tells the story and 190 00:12:46.320 --> 00:12:50.350 A:middle L:90% presents the challenges of what we're trying to do. 191 00:12:52.639 --> 00:12:56.840 A:middle L:90% We want on the order of 1000 fold increase in 192 00:12:56.840 --> 00:13:00.009 A:middle L:90% system performance. Well, we had a two pet 193 00:13:00.019 --> 00:13:01.860 A:middle L:90% slot machines so they chose 500 to put here. 194 00:13:03.139 --> 00:13:09.809 A:middle L:90% But power can increase by only a factor three memory 195 00:13:09.809 --> 00:13:15.179 A:middle L:90% band On the nose will increase by only a factor 196 00:13:15.190 --> 00:13:22.259 A:middle L:90% of 16 Storage capacity Will increase by only a factor 197 00:13:22.259 --> 00:13:26.919 A:middle L:90% of 20 and I. O. Bandwidth By only 198 00:13:26.919 --> 00:13:31.929 A:middle L:90% a factor of 100. Within these numbers comes a 199 00:13:31.929 --> 00:13:37.450 A:middle L:90% story in which flops are free. We no longer 200 00:13:37.450 --> 00:13:41.759 A:middle L:90% have to worry about the computational cost of a floating 201 00:13:41.759 --> 00:13:45.529 A:middle L:90% point operation. What we have to worry about and 202 00:13:45.539 --> 00:13:50.950 A:middle L:90% what we have to minimize his data movement. Mm 203 00:13:50.950 --> 00:13:54.830 A:middle L:90% . P A message passing interface is something that my 204 00:13:54.830 --> 00:13:58.559 A:middle L:90% management says we need to put a stake through its 205 00:13:58.559 --> 00:14:03.409 A:middle L:90% heart because it's so dependent on moving data around and 206 00:14:03.409 --> 00:14:07.059 A:middle L:90% you can't afford to do it. It's difficult enough 207 00:14:07.200 --> 00:14:11.409 A:middle L:90% to move data and expensive enough to move data across 208 00:14:11.409 --> 00:14:16.210 A:middle L:90% the chip. Moving it across. A supercomputer is 209 00:14:16.210 --> 00:14:20.440 A:middle L:90% prohibitively expensive in terms of data getting it off the 210 00:14:20.440 --> 00:14:28.070 A:middle L:90% machine is daunting for a variety of reasons. Basically 211 00:14:28.080 --> 00:14:35.289 A:middle L:90% if you listen to computational scientist this architecture he is 212 00:14:35.289 --> 00:14:39.019 A:middle L:90% hostile to science and it's up to us in the 213 00:14:39.019 --> 00:14:45.200 A:middle L:90% computer science research program to develop the software stack that 214 00:14:45.210 --> 00:14:50.279 A:middle L:90% will let scientific codes run in this environment. That 215 00:14:50.289 --> 00:14:54.149 A:middle L:90% at the hardware level doesn't look very friendly to scientists 216 00:14:58.539 --> 00:15:01.610 A:middle L:90% . Besides the challenges of data movement, we're looking 217 00:15:01.610 --> 00:15:03.950 A:middle L:90% at a machine with so many parts that hardware failures 218 00:15:03.960 --> 00:15:11.500 A:middle L:90% will be almost continuous because of reduced scale and running 219 00:15:11.509 --> 00:15:16.100 A:middle L:90% hardware at reduced voltages at threshold voltage levels. We'll 220 00:15:16.100 --> 00:15:20.279 A:middle L:90% be looking at soft errors that are not detected in 221 00:15:20.289 --> 00:15:24.700 A:middle L:90% hardware so that we have a machine that I'm not 222 00:15:24.700 --> 00:15:28.789 A:middle L:90% deterministic anymore. That gives us a host of challenges 223 00:15:28.789 --> 00:15:35.940 A:middle L:90% in the program ability and in uncertainty, quantification programming 224 00:15:35.950 --> 00:15:39.059 A:middle L:90% the scale of billion way concurrency is mind boggling to 225 00:15:39.059 --> 00:15:43.620 A:middle L:90% most people. You think of the challenges of getting 226 00:15:43.620 --> 00:15:50.470 A:middle L:90% someone to understand hundredweight concurrency to think about coordinating computation 227 00:15:50.470 --> 00:15:54.580 A:middle L:90% at that level and then think about billion Waken currency 228 00:15:54.590 --> 00:15:58.100 A:middle L:90% . And you will quickly realize we do not have 229 00:15:58.110 --> 00:16:03.649 A:middle L:90% the intellectual constructs to think about codes effectively at that 230 00:16:03.840 --> 00:16:07.940 A:middle L:90% . Imagine trying to debug a code on a non 231 00:16:07.940 --> 00:16:12.870 A:middle L:90% deterministic machine with that level of concurrency and think about 232 00:16:12.879 --> 00:16:18.250 A:middle L:90% the kinds of tools and technologies that we must develop 233 00:16:18.740 --> 00:16:22.490 A:middle L:90% so that these machines are easy to use. Yes 234 00:16:22.500 --> 00:16:26.549 A:middle L:90% , easy to use. No longer requiring only hero 235 00:16:26.549 --> 00:16:30.659 A:middle L:90% programmers. And how do we find that level of 236 00:16:30.669 --> 00:16:34.090 A:middle L:90% concurrency within a scientific application? This is a non 237 00:16:34.090 --> 00:16:42.769 A:middle L:90% trivial. From the scientist perspective, they have adapted 238 00:16:42.779 --> 00:16:47.960 A:middle L:90% to move into new machines thus far by simply scaling 239 00:16:47.960 --> 00:16:49.470 A:middle L:90% up the size of the problems that they attempted. 240 00:16:51.440 --> 00:16:55.740 A:middle L:90% But because this machine will have a smaller memory footprint 241 00:16:55.750 --> 00:16:59.450 A:middle L:90% relative to the size of computation, that's not going 242 00:16:59.450 --> 00:17:02.389 A:middle L:90% to be the way that we achieve billion away concurrency 243 00:17:02.399 --> 00:17:06.619 A:middle L:90% anymore. And the challenge that we face is that 244 00:17:06.619 --> 00:17:11.099 A:middle L:90% scientists much to completely reformulate their approaches to doing their 245 00:17:11.109 --> 00:17:15.380 A:middle L:90% physics, their chemistry, their biological modelling, their 246 00:17:15.380 --> 00:17:19.650 A:middle L:90% stockpile management. It is a huge challenge that we 247 00:17:19.660 --> 00:17:23.329 A:middle L:90% have begun to address. But we are at the 248 00:17:23.339 --> 00:17:32.509 A:middle L:90% very beginning of this great adventure. The implications of 249 00:17:32.519 --> 00:17:36.220 A:middle L:90% a solution to our exa scale challenges will reach all 250 00:17:36.220 --> 00:17:38.869 A:middle L:90% the way to your desktop. Because if we are 251 00:17:38.880 --> 00:17:44.559 A:middle L:90% fully successful, you will be looking at desktop systems 252 00:17:44.559 --> 00:17:48.630 A:middle L:90% that are at terrorists came And this department will likely 253 00:17:48.630 --> 00:17:52.839 A:middle L:90% own at least one penny scale system 10 to 15 254 00:17:52.839 --> 00:17:56.400 A:middle L:90% years from now. So you might begin thinking about 255 00:17:56.410 --> 00:17:59.750 A:middle L:90% how you would use such a resource. I would 256 00:17:59.750 --> 00:18:03.599 A:middle L:90% also suggest to you and I don't think I have 257 00:18:03.609 --> 00:18:06.019 A:middle L:90% to go on this rant with this department, but 258 00:18:06.019 --> 00:18:10.440 A:middle L:90% I do with some that it is unethical to graduate 259 00:18:10.440 --> 00:18:14.240 A:middle L:90% someone with a degree in computer science who doesn't understand 260 00:18:14.250 --> 00:18:18.220 A:middle L:90% parallel programming in a world where you can no longer 261 00:18:18.220 --> 00:18:23.559 A:middle L:90% purchase a unit processor machine. I'm sad to say 262 00:18:23.569 --> 00:18:27.960 A:middle L:90% there are many universities that need that ran and then 263 00:18:29.640 --> 00:18:36.220 A:middle L:90% like Mhm. Mhm. At the same time that 264 00:18:36.230 --> 00:18:38.809 A:middle L:90% we are addressing the challenges of Exa Scale, we 265 00:18:38.809 --> 00:18:44.269 A:middle L:90% are mindful. Exa Scale systems present their own big 266 00:18:44.279 --> 00:18:47.529 A:middle L:90% data challenges. When you talk about building way, 267 00:18:47.529 --> 00:18:51.210 A:middle L:90% concurrency and the data that results the rest of the 268 00:18:51.210 --> 00:18:55.220 A:middle L:90% Office of Science and building their new user facilities, 269 00:18:55.230 --> 00:18:59.720 A:middle L:90% alert us to their own challenges with data intensive science 270 00:19:00.029 --> 00:19:03.259 A:middle L:90% . And what the popular press has name big data 271 00:19:04.039 --> 00:19:08.319 A:middle L:90% . We're looking at genomics data at our joint genomics 272 00:19:08.319 --> 00:19:15.200 A:middle L:90% Institute on the scale of 10 petabytes By 2021 and 273 00:19:15.200 --> 00:19:19.049 A:middle L:90% hired high energy physics with the large Hadron collider. 274 00:19:19.440 --> 00:19:25.109 A:middle L:90% That's just one of the instruments for ACP. They're 275 00:19:25.109 --> 00:19:29.119 A:middle L:90% talking about 15 petabytes of data per year but they 276 00:19:29.119 --> 00:19:30.690 A:middle L:90% don't just throw the data the way they keep it 277 00:19:30.700 --> 00:19:34.730 A:middle L:90% for years and continue to compare and contrast. So 278 00:19:34.730 --> 00:19:40.809 A:middle L:90% that number increases. We operate light sources that produce 279 00:19:40.809 --> 00:19:47.160 A:middle L:90% 300 terabytes of data per day increasing continuously. And 280 00:19:47.160 --> 00:19:51.089 A:middle L:90% climate data, which is our largest growing data collection 281 00:19:51.089 --> 00:19:53.289 A:middle L:90% at most of our facilities will be in the range 282 00:19:53.289 --> 00:20:02.720 A:middle L:90% of multiple exabytes within a decade. As I will 283 00:20:02.730 --> 00:20:06.490 A:middle L:90% talk through some climate examples here. I'll do that 284 00:20:06.490 --> 00:20:10.470 A:middle L:90% with precaution. I'm not an expert in modeling and 285 00:20:10.470 --> 00:20:14.109 A:middle L:90% simulation and I'm not a climate scientist and I didn't 286 00:20:14.109 --> 00:20:18.519 A:middle L:90% even stay at a Holiday Inn Express last but I 287 00:20:18.519 --> 00:20:21.900 A:middle L:90% know some experts in this area pretty well. Could 288 00:20:21.900 --> 00:20:25.230 A:middle L:90% provide me with a lot of interesting material. And 289 00:20:25.230 --> 00:20:27.359 A:middle L:90% one of the things I'm going to show you is 290 00:20:27.359 --> 00:20:30.920 A:middle L:90% that when we talk about big data and data intensive 291 00:20:30.920 --> 00:20:37.309 A:middle L:90% science big is not necessarily the important word. It's 292 00:20:37.309 --> 00:20:41.349 A:middle L:90% not just petabytes or exabytes that we have to worry 293 00:20:41.339 --> 00:20:47.269 A:middle L:90% . It's data with 10,000 dimensions that we have to 294 00:20:47.279 --> 00:20:52.740 A:middle L:90% worry. It's complex systems that interact with one in 295 00:20:52.740 --> 00:20:56.730 A:middle L:90% many ways. So this graph is showing you the 296 00:20:56.730 --> 00:21:02.519 A:middle L:90% various options for how climate might evolve and what we 297 00:21:02.519 --> 00:21:06.940 A:middle L:90% believe natural forces would produce in terms of normal climate 298 00:21:06.940 --> 00:21:11.150 A:middle L:90% variability. That's the blue band and then there's what 299 00:21:11.150 --> 00:21:15.210 A:middle L:90% we believe would happen under continued forcing because our energy 300 00:21:15.210 --> 00:21:21.210 A:middle L:90% production methods and that's the pink on the black is 301 00:21:21.210 --> 00:21:26.160 A:middle L:90% the observed reality. There are those who deny that 302 00:21:26.170 --> 00:21:30.579 A:middle L:90% climate change is occurring and they deny that human behavior 303 00:21:30.579 --> 00:21:37.119 A:middle L:90% is causing energy production methods or causing. But recent 304 00:21:37.130 --> 00:21:41.269 A:middle L:90% articles in Science magazine in The New york Times point 305 00:21:41.269 --> 00:21:45.069 A:middle L:90% out that the most pessimistic of the climate models are 306 00:21:45.069 --> 00:21:51.950 A:middle L:90% the ones that are more accurately forecast recent trends in 307 00:21:51.950 --> 00:21:56.960 A:middle L:90% climate around the world. So why does deal we 308 00:21:56.960 --> 00:22:00.559 A:middle L:90% care? Let me give you just a few examples 309 00:22:02.640 --> 00:22:07.200 A:middle L:90% . We care about energy production. 80 of US 310 00:22:07.200 --> 00:22:12.210 A:middle L:90% . refinery capacity is less than 1.5 m above sea 311 00:22:12.210 --> 00:22:18.500 A:middle L:90% level, 80%. We have a great many power 312 00:22:18.500 --> 00:22:23.349 A:middle L:90% plants along the coast as well as our concerns for 313 00:22:23.349 --> 00:22:27.180 A:middle L:90% refinery capacity. So that brings us to a deep 314 00:22:27.180 --> 00:22:32.819 A:middle L:90% and abiding interest in sea level rise. We know 315 00:22:32.829 --> 00:22:36.890 A:middle L:90% , for example, that just just melting the Greenland 316 00:22:36.890 --> 00:22:40.769 A:middle L:90% ice sheet would raise sea levels by about 20 ft 317 00:22:41.640 --> 00:22:45.660 A:middle L:90% . The east arctic ice sheet, if melted entirely 318 00:22:47.240 --> 00:22:49.009 A:middle L:90% , Could raise sea level by as much as 200 319 00:22:49.019 --> 00:22:53.789 A:middle L:90% ft. Is that likely within this century, happily 320 00:22:53.799 --> 00:22:59.279 A:middle L:90% . No, there are some studies that say no 321 00:22:59.279 --> 00:23:00.799 A:middle L:90% , it's not likely that sea level will rise more 322 00:23:00.799 --> 00:23:07.720 A:middle L:90% than 6.5 ft or so in this century. That's 323 00:23:07.720 --> 00:23:10.839 A:middle L:90% enough to put the tidal basin in the Washington mall 324 00:23:10.839 --> 00:23:18.210 A:middle L:90% underwater. Maybe someone in Washington attention, but we're 325 00:23:18.210 --> 00:23:25.380 A:middle L:90% not counting on that. We know that in the 326 00:23:25.380 --> 00:23:29.910 A:middle L:90% last interglacial period sea level was 4-6 m above where 327 00:23:29.910 --> 00:23:33.470 A:middle L:90% it is now. We may be lucky if that's 328 00:23:33.470 --> 00:23:36.500 A:middle L:90% as far as it goes. I'm not going to 329 00:23:36.500 --> 00:23:38.190 A:middle L:90% spend a whole lot of time on this, giving 330 00:23:38.190 --> 00:23:42.140 A:middle L:90% you lots of pointers into reports because I hope you 331 00:23:42.140 --> 00:23:48.799 A:middle L:90% okay. Another reason to care. I have a 332 00:23:48.809 --> 00:23:51.690 A:middle L:90% degree from the University of New Orleans. So I 333 00:23:51.690 --> 00:23:56.299 A:middle L:90% look at this particular slide pumps out 2400 miles of 334 00:23:56.309 --> 00:24:00.880 A:middle L:90% highway Along the Gulf Coast, less than four ft 335 00:24:00.880 --> 00:24:06.319 A:middle L:90% above sea level. And we thought Katrina and rita 336 00:24:06.329 --> 00:24:10.619 A:middle L:90% did a number on New Orleans. Sea level rise 337 00:24:10.630 --> 00:24:17.569 A:middle L:90% made it worse. We're looking not only at sea 338 00:24:17.569 --> 00:24:22.049 A:middle L:90% level rise, but other implications of global change, 339 00:24:22.440 --> 00:24:26.460 A:middle L:90% changing patterns of precipitation, both in terms of flood 340 00:24:26.470 --> 00:24:30.259 A:middle L:90% and drought. More severe hurricanes, possibly more tornadoes 341 00:24:30.740 --> 00:24:34.269 A:middle L:90% . We're also looking at implications for the food supply 342 00:24:34.940 --> 00:24:40.250 A:middle L:90% , water energy, an impact on human health and 343 00:24:40.259 --> 00:24:48.880 A:middle L:90% behavior. by 2100, it is expected that temperatures 344 00:24:48.890 --> 00:24:53.450 A:middle L:90% will rise any quarter from 2 to 11.5 degrees Fahrenheit 345 00:24:53.460 --> 00:25:03.269 A:middle L:90% on average two 11.5 degrees. And remember my earlier 346 00:25:03.279 --> 00:25:06.700 A:middle L:90% comment that the more pessimistic bottles are the ones that 347 00:25:06.700 --> 00:25:10.539 A:middle L:90% are tending to be the most now. Yeah, 348 00:25:11.140 --> 00:25:15.079 A:middle L:90% very difficult things for us to manage in terms of 349 00:25:15.079 --> 00:25:18.440 A:middle L:90% rising temperatures and impact on health. Mhm. This 350 00:25:18.440 --> 00:25:22.779 A:middle L:90% would more than double the days over 90° for a 351 00:25:22.779 --> 00:25:25.859 A:middle L:90% large part of the United States by the end of 352 00:25:25.869 --> 00:25:30.809 A:middle L:90% this century. As shown in Iskra. It also 353 00:25:30.809 --> 00:25:34.970 A:middle L:90% has implications for food production as some of our major 354 00:25:34.970 --> 00:25:41.589 A:middle L:90% crops like corn and soy will cease to parliament When 355 00:25:41.589 --> 00:25:45.319 A:middle L:90% the temperature rises by two or 3°, there goes 356 00:25:45.329 --> 00:25:49.730 A:middle L:90% corn and soy at a time when population is increasing 357 00:25:49.740 --> 00:25:55.930 A:middle L:90% dramatically And we already expect to need a 50 increase 358 00:25:55.930 --> 00:25:57.670 A:middle L:90% in the food supply before the end of the century 359 00:26:02.440 --> 00:26:04.819 A:middle L:90% . In terms of availability of water, I think 360 00:26:04.819 --> 00:26:08.690 A:middle L:90% most people have seen that texas has been in persistent 361 00:26:08.690 --> 00:26:14.869 A:middle L:90% and severe drought for some time. The Mississippi waterway 362 00:26:14.869 --> 00:26:18.740 A:middle L:90% was so affected by drought this summer. Transportation on 363 00:26:18.740 --> 00:26:21.640 A:middle L:90% it stopped for a period of time because the water 364 00:26:21.640 --> 00:26:23.849 A:middle L:90% levels were too low. And yet if you're in 365 00:26:23.849 --> 00:26:26.890 A:middle L:90% the Northeast, it is likely to get cooler and 366 00:26:26.900 --> 00:26:33.579 A:middle L:90% wetter with implications for flooding. While the Southwest is 367 00:26:33.579 --> 00:26:37.599 A:middle L:90% looking at worsening drought and decreasing snowpack in the mountains 368 00:26:37.609 --> 00:26:44.710 A:middle L:90% , worsening drought situation for the summer. Meanwhile, 369 00:26:45.019 --> 00:26:48.400 A:middle L:90% like the brilliant population that we are, everyone in 370 00:26:48.400 --> 00:26:52.170 A:middle L:90% the United States wants to move to California, mhm 371 00:26:53.740 --> 00:26:57.759 A:middle L:90% . To the southwest to join the wars for water 372 00:26:59.440 --> 00:27:03.549 A:middle L:90% and the rationing and the burns longs and the brownouts 373 00:27:03.140 --> 00:27:06.450 A:middle L:90% . Because we're so smart about these things, right 374 00:27:07.339 --> 00:27:11.349 A:middle L:90% ? Um This is an indication of population movement. 375 00:27:11.740 --> 00:27:14.660 A:middle L:90% As you see the rural areas of the country where 376 00:27:14.660 --> 00:27:21.210 A:middle L:90% drought will be less likely losing population while the worst 377 00:27:21.220 --> 00:27:25.359 A:middle L:90% hit areas in terms of drought and diminishing water supply 378 00:27:25.740 --> 00:27:29.859 A:middle L:90% will be gaining population at a tremendous rate and the 379 00:27:29.859 --> 00:27:33.500 A:middle L:90% areas most likely to be affected by sea level rise 380 00:27:33.500 --> 00:27:37.640 A:middle L:90% , like florida remains a magnet for people who want 381 00:27:37.640 --> 00:27:42.970 A:middle L:90% to move there. One of the things that's likely 382 00:27:42.970 --> 00:27:45.710 A:middle L:90% to happen with climate change is that we become a 383 00:27:45.720 --> 00:27:53.859 A:middle L:90% warmer wetter planet in some places. Dryer by others 384 00:27:55.440 --> 00:27:57.539 A:middle L:90% . So you look in this slide at the increase 385 00:27:57.549 --> 00:28:03.529 A:middle L:90% in severe weather events and heavy precipitation events and look 386 00:28:03.529 --> 00:28:07.210 A:middle L:90% at what happens in the Northeast around new york city 387 00:28:07.420 --> 00:28:12.670 A:middle L:90% and boston, Where we have a 67 increase in 388 00:28:12.680 --> 00:28:18.559 A:middle L:90% heavy precipitation. What's talked about now is 100 year 389 00:28:18.559 --> 00:28:21.670 A:middle L:90% flood is likely to become more 10 years. For 390 00:28:23.539 --> 00:28:29.619 A:middle L:90% those are the individualization share my pain at this image 391 00:28:29.630 --> 00:28:33.150 A:middle L:90% , which is showing you the dark blue is water 392 00:28:33.180 --> 00:28:37.380 A:middle L:90% around the tip of Manhattan and quiet this building chill 393 00:28:37.390 --> 00:28:45.559 A:middle L:90% man and the pale blue is flood around the bill 394 00:28:47.140 --> 00:28:49.200 A:middle L:90% . This is famous 100 year flood zone for the 395 00:28:49.200 --> 00:28:55.019 A:middle L:90% city. If you think about what happened during Hurricane 396 00:28:55.019 --> 00:28:57.599 A:middle L:90% Sandy, the flooding of the subway system, the 397 00:28:57.609 --> 00:29:00.349 A:middle L:90% flooding in the rural areas around new york city, 398 00:29:02.140 --> 00:29:07.460 A:middle L:90% this is what we're and having a tough increasing frequency 399 00:29:08.339 --> 00:29:11.190 A:middle L:90% . Consider also that some of those buildings flooded are 400 00:29:11.190 --> 00:29:15.950 A:middle L:90% likely to be power facilities. We hope there are 401 00:29:15.440 --> 00:29:26.809 A:middle L:90% Nuclear generators Plus five in that context, Current Energy 402 00:29:26.809 --> 00:29:34.920 A:middle L:90% Secretary chu observes. We seek solutions. We don't 403 00:29:34.920 --> 00:29:41.450 A:middle L:90% see, dare I say this. Just scientific papers 404 00:29:42.140 --> 00:29:45.509 A:middle L:90% . We are looking for people who want to make 405 00:29:45.509 --> 00:29:52.769 A:middle L:90% a difference, who want not to admire problems but 406 00:29:52.740 --> 00:29:57.359 A:middle L:90% but to make a difference, we need help, 407 00:29:59.539 --> 00:30:03.119 A:middle L:90% we need a lot to help us with the modeling 408 00:30:03.119 --> 00:30:07.769 A:middle L:90% and simulation to help us understand the implications on health 409 00:30:07.140 --> 00:30:15.559 A:middle L:90% , communicate with policy makers find solutions to mitigate problems 410 00:30:15.839 --> 00:30:22.900 A:middle L:90% the climate change bring to society in terms of data 411 00:30:22.900 --> 00:30:26.789 A:middle L:90% science, this is a list of just you are 412 00:30:26.799 --> 00:30:33.170 A:middle L:90% the kinds of data that are involved understanding climate science 413 00:30:33.740 --> 00:30:37.400 A:middle L:90% and understanding its implications for society. We have data 414 00:30:37.400 --> 00:30:44.349 A:middle L:90% sources like mobile stations collecting observational data, high resolution 415 00:30:44.349 --> 00:30:48.230 A:middle L:90% weather stations, you've got full blown snow and weather 416 00:30:48.230 --> 00:30:53.319 A:middle L:90% stations satellite imagery, weather radar, you've got ocean 417 00:30:53.319 --> 00:31:00.089 A:middle L:90% current observations and air current observations, measurements of gas 418 00:31:00.089 --> 00:31:03.579 A:middle L:90% in the atmosphere and radiation in the atmosphere. You've 419 00:31:03.589 --> 00:31:07.730 A:middle L:90% got to deal with cyanosis, snow, hydrology potential 420 00:31:07.740 --> 00:31:11.779 A:middle L:90% for avalanches. You've got to understand the results from 421 00:31:11.779 --> 00:31:17.460 A:middle L:90% seismic probes. You've also got to understand things like 422 00:31:18.039 --> 00:31:23.400 A:middle L:90% urban infrastructure and population patterns to really assess what is 423 00:31:23.410 --> 00:31:27.470 A:middle L:90% happening in an era. At the Supercomputing conference in 424 00:31:27.470 --> 00:31:30.000 A:middle L:90% New Orleans a few years ago, the focus was 425 00:31:30.000 --> 00:31:33.670 A:middle L:90% on climate science and climate data. And I remember 426 00:31:33.670 --> 00:31:37.170 A:middle L:90% one of our leading climate scientists standing in front of 427 00:31:37.170 --> 00:31:41.730 A:middle L:90% an audience and talking about a great surprise. He 428 00:31:41.730 --> 00:31:45.960 A:middle L:90% and his colleagues had one night had completed a simulation 429 00:31:45.960 --> 00:31:48.779 A:middle L:90% and sea level rise and published a paper and the 430 00:31:48.789 --> 00:31:52.460 A:middle L:90% group of coastal city urban planners came and said, 431 00:31:53.640 --> 00:31:57.369 A:middle L:90% can we get your sea level rise data for our 432 00:31:57.380 --> 00:32:02.079 A:middle L:90% region because we want to figure out where we can 433 00:32:02.089 --> 00:32:07.750 A:middle L:90% put power plants and where we can locate new housing 434 00:32:07.759 --> 00:32:13.849 A:middle L:90% developments and have people be safe. And the climate 435 00:32:13.849 --> 00:32:19.809 A:middle L:90% scientists pause and thought about the problems that request presented 436 00:32:19.809 --> 00:32:22.839 A:middle L:90% to first was the fact that they hadn't kept the 437 00:32:22.849 --> 00:32:28.230 A:middle L:90% data once they had done their scientists. Then there 438 00:32:28.230 --> 00:32:30.109 A:middle L:90% was the small matter of the resolution of the models 439 00:32:30.109 --> 00:32:35.509 A:middle L:90% not being anywhere near precise enough to provide the kind 440 00:32:35.509 --> 00:32:39.200 A:middle L:90% of guidance that file this one. And there's the 441 00:32:39.210 --> 00:32:45.259 A:middle L:90% further challenge of something we're looking at very closely as 442 00:32:45.259 --> 00:32:49.700 A:middle L:90% we look at policy implications and that's that data collected 443 00:32:49.700 --> 00:32:53.710 A:middle L:90% in oneness are rarely captured with sufficient metadata and other 444 00:32:53.710 --> 00:32:59.250 A:middle L:90% forms of provenance to be understandable by people in another 445 00:32:59.250 --> 00:33:05.869 A:middle L:90% district. If you read the America competes reauthorization 2010 446 00:33:05.880 --> 00:33:08.240 A:middle L:90% will find in it. I believe it's on page 447 00:33:08.240 --> 00:33:14.890 A:middle L:90% 11 in section 13 A mandate to science agencies to 448 00:33:14.890 --> 00:33:20.700 A:middle L:90% begin discussion of policy around preservation of public access to 449 00:33:20.710 --> 00:33:25.660 A:middle L:90% data resulting from science that we support. In february 450 00:33:25.660 --> 00:33:29.369 A:middle L:90% of this year, the Office of Science and Technology 451 00:33:29.380 --> 00:33:32.910 A:middle L:90% Policy changed that recommendation to a mandate. Not only 452 00:33:32.910 --> 00:33:37.230 A:middle L:90% that we should discuss policy, but we should have 453 00:33:37.230 --> 00:33:45.309 A:middle L:90% an implementation plan for it within six months. The 454 00:33:45.319 --> 00:33:54.809 A:middle L:90% issues there are vast the implications are also fine for 455 00:33:54.809 --> 00:33:59.369 A:middle L:90% thousands of you interested in data analysis and visualization, 456 00:33:59.380 --> 00:34:04.970 A:middle L:90% this particular image. Mhm. You chose results from 457 00:34:04.970 --> 00:34:09.960 A:middle L:90% one multi marble cement ensembles on climate modeling and this 458 00:34:09.960 --> 00:34:16.550 A:middle L:90% shows clear evidence of global warming, atmospheric warming that 459 00:34:16.559 --> 00:34:23.260 A:middle L:90% results from human productivity of energy production and consumer energy 460 00:34:24.239 --> 00:34:28.500 A:middle L:90% . The upper air layers of the atmosphere that would 461 00:34:28.500 --> 00:34:31.260 A:middle L:90% normally be warmed by the sun are not. So 462 00:34:32.440 --> 00:34:36.389 A:middle L:90% it's at the lower mid range where you see the 463 00:34:36.389 --> 00:34:42.280 A:middle L:90% effects of greenhouse gases, see evidence of what some 464 00:34:42.280 --> 00:34:47.710 A:middle L:90% of us think it's pretty conclusive. Yes. So 465 00:34:47.710 --> 00:34:53.750 A:middle L:90% the challenge is data driven science requires high throughput very 466 00:34:53.750 --> 00:34:58.849 A:middle L:90% fast. I o. We've only recently within the 467 00:34:58.849 --> 00:35:02.250 A:middle L:90% last year learned that are super computers not only have 468 00:35:02.250 --> 00:35:07.550 A:middle L:90% an output bottleneck as we've known all along, but 469 00:35:07.559 --> 00:35:09.940 A:middle L:90% for many of our largest codes that use the largest 470 00:35:09.949 --> 00:35:14.670 A:middle L:90% numbers and I'm meaning billions of cycles on supercomputers, 471 00:35:15.239 --> 00:35:19.099 A:middle L:90% their input back. If you do graphs and their 472 00:35:19.110 --> 00:35:22.380 A:middle L:90% input and output, the input bars go off the 473 00:35:22.389 --> 00:35:24.610 A:middle L:90% graph and the output bars are invisible because they do 474 00:35:24.619 --> 00:35:30.590 A:middle L:90% relatively small amounts of out in comparison to the amount 475 00:35:30.590 --> 00:35:34.110 A:middle L:90% of data they input over the rod part of that 476 00:35:34.110 --> 00:35:37.909 A:middle L:90% I suspect results from our current checkpoint restart model of 477 00:35:37.909 --> 00:35:42.019 A:middle L:90% saving the machine states period so that when you have 478 00:35:42.019 --> 00:35:45.099 A:middle L:90% a hard fault input the day again and restart and 479 00:35:45.099 --> 00:35:49.789 A:middle L:90% I could speculate that system was experiencing a lot of 480 00:35:49.800 --> 00:35:53.070 A:middle L:90% false during that. But it's also increasingly the case 481 00:35:53.369 --> 00:35:59.269 A:middle L:90% that models are fed by experimental and observational data as 482 00:35:59.269 --> 00:36:05.239 A:middle L:90% a starting and that we need to compare and cross 483 00:36:05.239 --> 00:36:08.849 A:middle L:90% analyzed between simulation data and observation. Experimental data. 484 00:36:09.829 --> 00:36:15.030 A:middle L:90% Typically our analytic methods require large memory footprint with the 485 00:36:15.030 --> 00:36:21.000 A:middle L:90% excess feel architecture trying to provide that. So we 486 00:36:21.000 --> 00:36:24.460 A:middle L:90% know that for a variety of systems, both light 487 00:36:24.460 --> 00:36:30.139 A:middle L:90% sources climate systems simulations, we need to find better 488 00:36:30.139 --> 00:36:35.340 A:middle L:90% ways to deal with streaming data. Our current methods 489 00:36:35.340 --> 00:36:38.780 A:middle L:90% of analysing streaming data are tied up to hypothesis. 490 00:36:39.070 --> 00:36:43.869 A:middle L:90% They're biased in favor of proving or disproving a particular 491 00:36:43.880 --> 00:36:46.940 A:middle L:90% hypothesis. But that means that as we throw away 492 00:36:46.949 --> 00:36:50.699 A:middle L:90% a very, very high percentage of data from the 493 00:36:50.699 --> 00:36:53.820 A:middle L:90% experiment, we also throw away the opportunity to stuff 494 00:36:53.829 --> 00:36:59.750 A:middle L:90% to discover the unexpected that is less and less acceptable 495 00:37:00.090 --> 00:37:02.809 A:middle L:90% as experiments are expensive to run and many of them 496 00:37:02.809 --> 00:37:10.800 A:middle L:90% will never be run again. Our scientists spend way 497 00:37:10.800 --> 00:37:16.889 A:middle L:90% too much time studying machine architectures trying to learn enough 498 00:37:16.900 --> 00:37:21.179 A:middle L:90% computer science to work on large scale machines that are 499 00:37:21.190 --> 00:37:25.440 A:middle L:90% difficult for senior computer scientists understand and they're pulled away 500 00:37:25.440 --> 00:37:30.599 A:middle L:90% from their physics chemistry, their combustion science, we 501 00:37:30.599 --> 00:37:32.429 A:middle L:90% need to change that. We need to change the 502 00:37:32.429 --> 00:37:36.179 A:middle L:90% amount of time that they have to invest in managing 503 00:37:36.190 --> 00:37:38.550 A:middle L:90% data. And we need to give them better resources 504 00:37:38.929 --> 00:37:44.449 A:middle L:90% to understand their data and to extract from that date 505 00:37:45.130 --> 00:37:47.889 A:middle L:90% . The high consequence science that is important to our 506 00:37:47.889 --> 00:37:52.960 A:middle L:90% nation's future. We need ways to integrate data, 507 00:37:53.429 --> 00:37:58.639 A:middle L:90% we need ways to capture providence. Yeah. We 508 00:37:58.639 --> 00:38:00.630 A:middle L:90% need ways to make data usable. Not only by 509 00:38:00.630 --> 00:38:05.309 A:middle L:90% those who originally produced it but by others who need 510 00:38:05.309 --> 00:38:09.219 A:middle L:90% to use that data in combinations not originally conceived to 511 00:38:09.219 --> 00:38:14.119 A:middle L:90% solve problems, not within the thoughts of the scientists 512 00:38:14.119 --> 00:38:19.579 A:middle L:90% who created the data. Part of that is capturing 513 00:38:19.590 --> 00:38:23.239 A:middle L:90% day semantics and developing others were metadata within and across 514 00:38:23.239 --> 00:38:30.210 A:middle L:90% communities and helping people make sense of that metadata Human 515 00:38:30.219 --> 00:38:35.789 A:middle L:90% understand as well as a machine understand they need to 516 00:38:35.800 --> 00:38:37.449 A:middle L:90% be able to understand what's been done to the data 517 00:38:38.119 --> 00:38:40.360 A:middle L:90% . How has it been selected? How has it 518 00:38:40.369 --> 00:38:46.179 A:middle L:90% been normalized? How has it been downsized, compressed 519 00:38:46.300 --> 00:38:52.630 A:middle L:90% reformulate? The transformations are numerous and of those trivial 520 00:38:52.639 --> 00:38:55.489 A:middle L:90% things from multi scale of physics. Which scale does 521 00:38:55.489 --> 00:38:58.340 A:middle L:90% it happen? To be honest. We looked at 522 00:38:58.340 --> 00:39:02.570 A:middle L:90% scales that range from nano scale to planetary scales and 523 00:39:02.570 --> 00:39:08.440 A:middle L:90% time ranges from nanoseconds. Two centuries. We've been 524 00:39:08.440 --> 00:39:14.449 A:middle L:90% holding workshops with other offices bes as basic energy sciences 525 00:39:15.019 --> 00:39:16.809 A:middle L:90% . Right now, the model has been for scientists 526 00:39:16.809 --> 00:39:20.239 A:middle L:90% to take their data home on a thumb drive. 527 00:39:21.820 --> 00:39:23.650 A:middle L:90% Taking home a petabytes of data is somewhat more of 528 00:39:23.650 --> 00:39:27.889 A:middle L:90% a challenge. Most of us think that trying to 529 00:39:28.610 --> 00:39:32.989 A:middle L:90% get a bite to a laptop is called. So 530 00:39:32.989 --> 00:39:36.670 A:middle L:90% we have to help them find new models for doing 531 00:39:36.670 --> 00:39:40.539 A:middle L:90% science and doing science with collaborations that spend not only 532 00:39:40.539 --> 00:39:46.650 A:middle L:90% multiple institutions but sometimes multiple nations and multiple. This 533 00:39:51.320 --> 00:39:54.960 A:middle L:90% research challenges just a few album captured here how we 534 00:39:54.960 --> 00:40:00.449 A:middle L:90% manage move and analyzed data, how we transform data 535 00:40:00.449 --> 00:40:05.429 A:middle L:90% into knowledge. Because I would argue that if what 536 00:40:05.440 --> 00:40:08.800 A:middle L:90% comes off of a supercomputer believed inside you might as 537 00:40:08.800 --> 00:40:12.820 A:middle L:90% well not have turned the large hot water heater on 538 00:40:14.809 --> 00:40:17.429 A:middle L:90% . We need ways to define metadata and methods to 539 00:40:17.440 --> 00:40:22.639 A:middle L:90% generate it automatically and capture it. We need new 540 00:40:22.639 --> 00:40:28.050 A:middle L:90% scientific work clothes systems sophisticated about the challenges of science 541 00:40:28.610 --> 00:40:31.469 A:middle L:90% and about the machine architectures they run a we need 542 00:40:31.469 --> 00:40:36.880 A:middle L:90% to understand the strengths and weaknesses of standards and we 543 00:40:36.880 --> 00:40:40.849 A:middle L:90% need to understand how to better communicate the scientific findings 544 00:40:40.849 --> 00:40:45.809 A:middle L:90% , not only to the scientists to the public and 545 00:40:45.809 --> 00:40:51.940 A:middle L:90% to policymakers and politicians. This okay. Mhm. 546 00:40:52.110 --> 00:40:57.789 A:middle L:90% We know that because of the changing supercomputer architecture we 547 00:40:57.789 --> 00:41:01.500 A:middle L:90% have to start from zero and build a whole new 548 00:41:01.510 --> 00:41:07.289 A:middle L:90% generation. Those on technologies. My office has supported 549 00:41:07.289 --> 00:41:14.210 A:middle L:90% development visit and power view to most large widely used 550 00:41:14.219 --> 00:41:17.860 A:middle L:90% scientific visualization and analysis tools and they're not going to 551 00:41:17.869 --> 00:41:23.530 A:middle L:90% run. And the excess scale environment without major reformulation 552 00:41:24.809 --> 00:41:30.159 A:middle L:90% . I've listened to ballrooms full of mathematicians, listen 553 00:41:30.159 --> 00:41:32.489 A:middle L:90% to destructions of this architecture, go away and working 554 00:41:34.210 --> 00:41:38.269 A:middle L:90% and come back and look at us baffled and angry 555 00:41:38.280 --> 00:41:40.739 A:middle L:90% and say the solvers are not going to run. 556 00:41:42.110 --> 00:41:50.139 A:middle L:90% Mr sure no more land. We have a problem 557 00:41:50.809 --> 00:41:54.280 A:middle L:90% we have to reformulate the solvers. We need new 558 00:41:54.289 --> 00:42:00.019 A:middle L:90% data structures and traversal out is that minimize data moves 559 00:42:00.300 --> 00:42:04.630 A:middle L:90% . Resilient in the face of hard and soft, 560 00:42:04.639 --> 00:42:09.369 A:middle L:90% false. We have a lot of work and I 561 00:42:09.369 --> 00:42:20.809 A:middle L:90% hope you'll join us. Thank you so much. 562 00:42:25.500 --> 00:42:30.619 A:middle L:90% Any questions I have just a little sense at the 563 00:42:30.630 --> 00:42:38.710 A:middle L:90% time there is a problem. Yeah, interesting is 564 00:42:38.719 --> 00:42:43.989 A:middle L:90% trying to present. I'm thinking about training students to 565 00:42:43.989 --> 00:42:46.219 A:middle L:90% participate in this. Well if you're here to take 566 00:42:46.219 --> 00:42:53.730 A:middle L:90% a position the need for new degrees um additional science 567 00:42:54.599 --> 00:43:01.289 A:middle L:90% darling. Right. one of the things that my 568 00:43:01.289 --> 00:43:08.480 A:middle L:90% office operators right now, yes. Yeah. Is 569 00:43:08.480 --> 00:43:14.420 A:middle L:90% a computational science graduate. All right. That support 570 00:43:15.099 --> 00:43:17.150 A:middle L:90% diverse disciplines. And one of the things that requires 571 00:43:17.460 --> 00:43:22.610 A:middle L:90% in exchange for years of graduate school support is that 572 00:43:22.610 --> 00:43:25.239 A:middle L:90% they take forces a variety of courses in computer science 573 00:43:25.250 --> 00:43:29.449 A:middle L:90% and math. And they also do internships at the 574 00:43:29.449 --> 00:43:36.510 A:middle L:90% national lab working on real science problems in resource environment 575 00:43:37.500 --> 00:43:42.460 A:middle L:90% . We're delighted six senator csg fellas go on to 576 00:43:42.469 --> 00:43:50.389 A:middle L:90% careers lands. What I'm here over from industry is 577 00:43:50.389 --> 00:43:52.460 A:middle L:90% that we don't turn out nearly enough people to understand 578 00:43:52.460 --> 00:43:57.360 A:middle L:90% why what we don't turn up. Nearly enough people 579 00:43:57.360 --> 00:44:00.280 A:middle L:90% to understand parallel program. I sat in a meeting 580 00:44:00.289 --> 00:44:06.869 A:middle L:90% of the Council on Competitiveness working with the interagency, 581 00:44:06.880 --> 00:44:09.110 A:middle L:90% working high end computing that on our co chair. 582 00:44:09.590 --> 00:44:13.750 A:middle L:90% And they were telling us these industry leaders from companies 583 00:44:13.750 --> 00:44:20.420 A:middle L:90% like GE were telling us that degrees may become irrelevant 584 00:44:20.989 --> 00:44:23.090 A:middle L:90% if the people involved. Those degrees don't have the 585 00:44:23.099 --> 00:44:28.429 A:middle L:90% skills that industry needs right now. The biggest barrier 586 00:44:28.440 --> 00:44:31.250 A:middle L:90% to their use of modeling and simulation is the inability 587 00:44:31.250 --> 00:44:36.510 A:middle L:90% to hire know how to do parallel programming here. 588 00:44:36.510 --> 00:44:42.699 A:middle L:90% My land. Mm You we need people do not 589 00:44:42.710 --> 00:44:45.550 A:middle L:90% have a clinton scale. We need people who have 590 00:44:45.550 --> 00:44:52.219 A:middle L:90% the interdisciplinary understanding to work with scientists understand the science 591 00:44:52.389 --> 00:44:57.619 A:middle L:90% and to work within these large scale environments to solve 592 00:44:57.619 --> 00:45:00.550 A:middle L:90% the important problems whether their problems of climate change or 593 00:45:00.559 --> 00:45:05.099 A:middle L:90% homeland security. Where data challenges are at least as 594 00:45:05.099 --> 00:45:12.510 A:middle L:90% a we are starving in the national lab and part 595 00:45:13.190 --> 00:45:15.820 A:middle L:90% for people with the skills that we need. And 596 00:45:15.820 --> 00:45:19.559 A:middle L:90% then few people that we have, particularly with data 597 00:45:19.559 --> 00:45:22.309 A:middle L:90% science skills are being sucked up by industry so fast 598 00:45:22.789 --> 00:45:27.070 A:middle L:90% that I wonder how we're going to achieve the scientific 599 00:45:27.070 --> 00:45:30.420 A:middle L:90% solution. But life form will find we depend. 600 00:45:31.190 --> 00:45:58.800 A:middle L:90% Yes, we need more of this. Mhm. 601 00:45:59.190 --> 00:46:10.940 A:middle L:90% Yeah. Yeah. Higher the nation. Lot of 602 00:46:10.940 --> 00:46:17.039 A:middle L:90% your slides mm offering now exactly those slides. Most 603 00:46:17.039 --> 00:46:22.300 A:middle L:90% of those images come from the global plant. Yeah 604 00:46:22.780 --> 00:46:24.840 A:middle L:90% . Global change dot gov site. They're not my 605 00:46:24.849 --> 00:46:30.199 A:middle L:90% creations. I'm an artist from some of those choices 606 00:46:30.880 --> 00:46:35.079 A:middle L:90% . Lightning wins. As I said, I'm one 607 00:46:35.079 --> 00:46:37.170 A:middle L:90% of the co chairs of the Interagency Working Group on 608 00:46:37.179 --> 00:46:42.780 A:middle L:90% high end Computing. That's part of the national coordination 609 00:46:42.780 --> 00:46:46.199 A:middle L:90% office for networking and Research and Development. We have 610 00:46:46.210 --> 00:46:52.570 A:middle L:90% regular meetings on multiple topics that bring together laughter, 611 00:46:52.739 --> 00:46:55.789 A:middle L:90% no different parts of the D. O. D 612 00:46:55.800 --> 00:47:00.110 A:middle L:90% . S. L. Any agency that is doing 613 00:47:00.119 --> 00:47:02.570 A:middle L:90% work in a particular area is likely to have a 614 00:47:02.579 --> 00:47:07.090 A:middle L:90% representative meetings. When you talk about data, quickly 615 00:47:07.090 --> 00:47:13.989 A:middle L:90% add anyone from veterans affairs to NIH to Homeland Security 616 00:47:14.679 --> 00:47:19.260 A:middle L:90% . So yes there is a lot inter agency coordination 617 00:47:19.260 --> 00:47:22.989 A:middle L:90% and discuss. one of the things under discussion in 618 00:47:22.480 --> 00:47:28.219 A:middle L:90% the hit group right now is developing a new federal 619 00:47:28.219 --> 00:47:31.920 A:middle L:90% plan for high end computing, Observing to our embarrassment 620 00:47:31.929 --> 00:47:38.010 A:middle L:90% that the one currently on the website has 2004 and 621 00:47:38.010 --> 00:47:42.590 A:middle L:90% is aimed at developing scale systems. We got there 622 00:47:45.480 --> 00:47:50.679 A:middle L:90% other questions. Yes. Mhm. Yeah. Actually 623 00:47:50.690 --> 00:47:53.199 A:middle L:90% thank you for a great time. Great. So 624 00:47:54.880 --> 00:47:59.480 A:middle L:90% you see quantum computing helping out with your mission and 625 00:47:59.480 --> 00:48:02.019 A:middle L:90% your order is your office looking at quantum computing at 626 00:48:02.019 --> 00:48:07.050 A:middle L:90% all parts of the Office of Science. Look at 627 00:48:07.059 --> 00:48:10.809 A:middle L:90% quantum computing. For the most part computing research is 628 00:48:10.809 --> 00:48:15.760 A:middle L:90% done by other agencies. I Arka has an interest 629 00:48:15.760 --> 00:48:16.989 A:middle L:90% in it and DARPA has an interest in it. 630 00:48:17.570 --> 00:48:21.849 A:middle L:90% I don't know what the NSL still does, but 631 00:48:21.849 --> 00:48:25.500 A:middle L:90% again, remember that we come together for discussions regularly 632 00:48:27.670 --> 00:48:31.219 A:middle L:90% as a mission agency. We don't have the luxury 633 00:48:31.219 --> 00:48:35.400 A:middle L:90% of doing science for the sake of science or computer 634 00:48:35.400 --> 00:48:37.550 A:middle L:90% science. For the sake of computer science. We 635 00:48:37.550 --> 00:48:40.360 A:middle L:90% have to keep our eye on the goal and the 636 00:48:40.369 --> 00:48:45.340 A:middle L:90% goal for us is supporting the science that's essential to 637 00:48:45.349 --> 00:48:49.650 A:middle L:90% the nation. With respect to energy. We have 638 00:48:49.650 --> 00:48:53.099 A:middle L:90% to be able to run the nuclear weapons stockpile simulation 639 00:48:53.099 --> 00:48:57.960 A:middle L:90% codes to support the annual certifications to the President and 640 00:48:57.960 --> 00:49:00.090 A:middle L:90% the Congress. We have to be able to do 641 00:49:00.469 --> 00:49:05.159 A:middle L:90% the research and combustion that we must support to reach 642 00:49:05.159 --> 00:49:07.579 A:middle L:90% some of the nation's goals With regard to reducing emissions 643 00:49:08.070 --> 00:49:12.639 A:middle L:90% . We have to support the climate research and related 644 00:49:12.639 --> 00:49:15.820 A:middle L:90% modeling and simulation and that means that we are somewhat 645 00:49:15.829 --> 00:49:22.000 A:middle L:90% conservative in our choices of architectures because we have millions 646 00:49:22.000 --> 00:49:28.010 A:middle L:90% and millions and millions of lines of science code that 647 00:49:28.010 --> 00:49:31.559 A:middle L:90% must continue to run and we can't afford to rewrite 648 00:49:31.559 --> 00:49:36.400 A:middle L:90% it all from scratch for a new architecture. It's 649 00:49:36.400 --> 00:49:39.570 A:middle L:90% hard enough to think about an exact scale architecture and 650 00:49:39.570 --> 00:49:44.300 A:middle L:90% how those legacy codes will continue to run those of 651 00:49:44.300 --> 00:49:46.860 A:middle L:90% you in software engineering. Do the quick math. 652 00:49:47.030 --> 00:49:51.730 A:middle L:90% What is it still around$100$110 per line of 653 00:49:51.739 --> 00:49:55.829 A:middle L:90% good quality code at millions and millions and millions of 654 00:49:55.829 --> 00:50:00.179 A:middle L:90% lines of code. We can't we can't afford to 655 00:50:00.179 --> 00:50:02.099 A:middle L:90% rewrite it. And in the case of some of 656 00:50:02.099 --> 00:50:07.360 A:middle L:90% those codes, the scientists who would be needed to 657 00:50:07.369 --> 00:50:13.039 A:middle L:90% validate the codes are no longer with us. We 658 00:50:13.039 --> 00:50:15.280 A:middle L:90% don't have a lot of nuclear weapons desires around. 659 00:50:16.659 --> 00:50:21.780 A:middle L:90% So messing with the codes that they have validated is 660 00:50:21.780 --> 00:50:29.690 A:middle L:90% not a good career. Other questions. Yes. 661 00:50:34.659 --> 00:50:39.750 A:middle L:90% So the question is basically Okay. Region solving the 662 00:50:39.760 --> 00:50:45.170 A:middle L:90% problem. Getting complications but solve the problem. Is 663 00:50:45.170 --> 00:50:50.070 A:middle L:90% that right? So in terms of solving problems happens 664 00:50:50.659 --> 00:50:55.059 A:middle L:90% lots of industry. Thanks solutions of islam. Yeah 665 00:50:57.860 --> 00:51:00.030 A:middle L:90% . You comment on um he has some kind of 666 00:51:00.039 --> 00:51:07.929 A:middle L:90% values have encouraged candy and in operations that winds and 667 00:51:07.940 --> 00:51:16.030 A:middle L:90% daniel. Mhm. Certainly. And our excess health 668 00:51:16.030 --> 00:51:22.869 A:middle L:90% program funds people in industry along with people in academia 669 00:51:22.880 --> 00:51:27.500 A:middle L:90% and the national labs. I can't think of any 670 00:51:27.500 --> 00:51:30.659 A:middle L:90% of the programs of research that we operate that are 671 00:51:30.699 --> 00:51:34.969 A:middle L:90% willing to fund industry partners in collaboration. In fact 672 00:51:34.980 --> 00:51:39.190 A:middle L:90% , for our latest X stacked solicitation that's focused on 673 00:51:39.760 --> 00:51:46.119 A:middle L:90% environment and execution models. I believe Intel is one 674 00:51:46.119 --> 00:51:52.289 A:middle L:90% of the leading organizations. They're not a sub awardee 675 00:51:52.289 --> 00:51:55.559 A:middle L:90% . They're the primary on that award. Similarly in 676 00:51:55.559 --> 00:52:00.210 A:middle L:90% some of our visualization codes we have small companies maintain 677 00:52:00.210 --> 00:52:05.219 A:middle L:90% those codes over and we happily support. We also 678 00:52:05.219 --> 00:52:07.980 A:middle L:90% are able to fund people in other countries and we 679 00:52:08.150 --> 00:52:12.489 A:middle L:90% we want to fund the best minds to solve our 680 00:52:12.489 --> 00:52:20.789 A:middle L:90% important problems wherever those minds are. Oh, let 681 00:52:20.789 --> 00:52:23.309 A:middle L:90% me remember, toss out another carrot for the faculty 682 00:52:23.309 --> 00:52:27.840 A:middle L:90% in the room. We don't have two weeks in 683 00:52:27.840 --> 00:52:32.429 A:middle L:90% the summer restriction of funding for faculty. We will 684 00:52:32.440 --> 00:52:37.989 A:middle L:90% have the loop on you for the academic year and 685 00:52:37.000 --> 00:52:40.840 A:middle L:90% we will happily support research As well as 10ure track 686 00:52:43.050 --> 00:52:46.929 A:middle L:90% and your graduate students. Also, Can I send 687 00:52:46.929 --> 00:52:51.980 A:middle L:90% you a sovereign song and get you away from NSF 688 00:52:52.750 --> 00:52:59.730 A:middle L:90% ? Mm Shit again. Thank our speaker for coming 689 00:52:59.739 --> 00:53:07.940 A:middle L:90% today, remind you all that there is a need 690 00:53:07.949 --> 00:53:14.730 A:middle L:90% to speaker section help for 5 15 and target. 691 00:53:14.829 --> 00:53:16.380 A:middle L:90% I'm sorry Mcbride, one of six this afternoon. 692 00:53:16.750 --> 00:53:22.269 A:middle L:90% Graduate students and undergraduates are all invited. Okay.