Building the Universe Pixel by Pixel

By Kelen Tuttle

August 20, 2014

Recently, the Harvard-Smithsonian Center for Astrophysics unveiled an unprecedented simulation of the universe’s development. Called the Illustris project, the simulation depicts more than 13 billion years of cosmic evolution across a cube of the universe that’s 350-million-light-years on each side. The goal was to view the formation of galaxies and other large-scale structure we see around us today, to test our understanding of what makes up the universe – including dark matter and dark energy – as well as how those components interact. It was a massive undertaking, one that took more than 5 years to complete. But why was it important to conduct such a simulation?

To better understand the science and art of astrophysics visualizations, three experts came together in late July to discuss the processes the ways in which their work benefits both science and the public’s perception of science. The participants:

RALF KAEHLER – is a physicist and computer scientist by training who now runs the visualization facilities at the Kavli Institute for Particle Astrophysics and Cosmology, located at SLAC National Accelerator Laboratory and Stanford University.

STUART LEVY – is a research programmer and member of the National Center for Supercomputing Applications’ Advanced Visualization Lab team, which creates high-resolution data-driven scientific visualizations for public outreach.

DYLAN NELSON – is a graduate student at the Harvard-Smithsonian Center for Astrophysics and a member of the Illustris collaboration, which recently completed a large cosmological simulation of galaxy formation.

The following is an edited transcript of a roundtable discussion. The participants have been provided the opportunity to amend or edit their remarks.

THE KAVLI FOUNDATION: Dylan, youre a member of the Illustris project team, so let’s start with you. Illustris was a massive undertaking, one that took more than 5 years to complete. Why was it important to conduct this simulation?

DYLAN NELSON: This simulation tested our big-picture understanding of the universe’s evolution. We can’t just compare the observation of one galaxy; we need to compare whole populations of thousands or tens of thousands of galaxies, and the simulation lets us do this by creating a big volume of the universe. In visualizing this simulation, we found some unexpected features that in retrospect really shouldn’t have been unexpected at all. For example, when we made a movie showing the temperature of gas in the universe evolving over time, we saw that galaxies had a tendency to flicker rapidly. We traced this back to one of the three ways in which we let supermassive black holes input energy into the galaxies within which they reside. Although we expected that the energy would affect the temperature of the gas, we didn’t know how intermittent it would be, how it would create these flickers and bursts. That’s really something that we were surprised by, and something that made us rework our models a bit. 

TKF: Ralf, what types of insights are gained through the visualizations you create with scientists at the Kavli Institute for Particle Astrophysics and Cosmology? Do you also tend to find unexpected features?

RALF KAEHLER: What I often hear from scientists is that they gain intuition from watching the animations we create, intuition that’s hard to get from just looking at the raw numbers. They see how gas moves, how dark matter clumps on smaller scales and then merges and forms larger and larger clumps of dark matter. And it seems like this intuition is very important for a thorough understanding of the processes.

Another very important advantage visualizations offer is the ability to catch errors in the simulations. By just looking at the numbers, it can be easy to miss these errors. But when watching an animation they can become totally obvious. We can easily see if there’s some discontinuity in the data that shouldn’t be there and then you can investigate further if it’s a feature or an artifact or a bug. The software used to produce these simulations usually consists of hundred thousands or millions of lines of code. Codes of that size often contain bugs and visualizations can help to determine if there’s an error hidden within the code.

TKF: It sounds like visualizations are especially good for identifying issues with your assumptions or the underlying models. Stuart, would you agree with that?

STUART LEVY: I think that’s a really good point, and it’s something that people talk about when they’re thinking of doing a visualization. If you reduce things to a graph with some statistics on it, in choosing what the statistics should measure, you’re saying what the interesting things are. And the hope is that if you can present something visually, you might end up bringing in things that you didn’t expect to bring in.

To me, it also seems like visualizations are becoming more and more useful for looking at very large-scale phenomena. As in observational astrophysics, instead of spending a lot of time looking at modest numbers of individual objects, people are looking at huge numbers of objects.

DYLAN NELSON: I agree with that. Large simulations like Illustris are similar to big observational surveys. When you’re not looking at individual objects, you need sophisticated visualization techniques to pull out the interesting information. For instance, back when the kind of cosmological simulations we do today first started, people plotted a point for each dark matter particle. They learned lots of science from doing that. But these days, when the biggest dark matter simulations include a trillion particles, that’s not going to get you as far. You’re going to need more sophisticated visualization approaches – as well as machine learning techniques or other automated ways of finding interesting trends in the simulation. We’re working on that.

TKF: Even though all three of you create visualizations, your roles and your connections to the scientific questions driving the research are different. How does the process work for each of you? Who comes up with the scientific questions you seek to answer?

RALF KAEHLER:

For us, it’s often an interactive process. We sit together in front of the screen and analyze the data in real-time. I try to design a lot of the algorithms in a way that they produce visualizations pretty quickly, so that we can change parameters like the camera position or the color maps in real time and get an updated image in a fraction of a second. That way, we can explore the data together, focusing on regions of interest, zooming in and out, things like that. Other times, it’s more offline, where I’ll render something overnight and send the result to the scientists and let them have a look at that.

I would also say that while half of my work is for scientists, the other half is for outreach. Sometimes we create visualizations purely for outreach purposes, and sometimes we can use the same visualization for both science and outreach. In the latter case, the scientists first analyze the dataset and then we tweak it a little bit, spending more time with the camera path and the color scheme to make it look a little bit prettier before we use it for planetarium shows.

DYLAN NELSON: The process for me is a little different because my primary responsibility is science. It’s only a secondary responsibility that lets me create visualizations. I always say that when I create visualizations, it’s both for scientific exploration and for dissemination to the public. But I think in reality, in my research group we do those two things in completely different ways.

We need visualizations to understand what is going on in a simulation, to better understand our models and the physical processes we’re simulating. But those visualizations are not pretty; we do them as quickly as possible, and as soon as we have a useful science result, the effort on the visualization stops. On the other hand, when we’re doing a visualization for outreach, that’s really intended to make people say “Oh, wow, that’s really cool!” So there’s a lot more time spent past the point of scientific realization, polishing and making the visualization look visually impressive. 

STUART LEVY: My group really focuses on outreach. So we usually have an idea for a show first, then we’ll go and look for scientists who work in that area and can provide the simulation. They’ll also tell us what we should believe from their simulations and what we shouldn’t believe. Often they’ll be making simulations they know are representing some aspects of reality well and others less well. And so they’ll say something like, don’t pay attention to the temperature here, since we’re not including everything that could be heating things up. We’ll go back and forth both with the scientists and the people producing the show to create something that’s both interesting and scientifically correct.

That said, we do occasionally work with scientists on unanswered questions – though it’s not always in the realm of astrophysics. A few years ago, we were working with a simulation of a tornado. One of the things that the scientists were interested in learning was the origin of tornados. Most severe storms don’t create tornados, so what’s special about the subset of storms that do? They had an idea that we should be looking in the simulation for a feature that’s called a rear flank downdraft – a storm that’s flowing in a certain way. We were looking for this signature in the visualizations and just not finding it. But then one of the graduate students picked out this sort of rolling feature – a horizontal bunch of rolling air – and succeeded in convincing his senior professors that, in this simulation at least, it was that feature and not a rear flank downdraft that triggered the tornado. That was a surprising result, one made possible by the visualization.

TKF: What have been the big breakthroughs in visualization in the past five years? Are there new technologies or revelations that make possible all youve just described? 

STUART LEVY: Bigger disks! It seems a little mundane, but the ability to store huge amounts of data is really important. A couple of years ago, we got about four terabytes of data from a scientist. A few years earlier, that would have been an overwhelming amount, but today we could easily take on several of those. That makes a really big difference. The billion-dollar gaming industry has also been an incredible boon to us. It’s on the back of that industry that high performance graphics cards have been built. Fifteen years ago, the fastest graphics hardware cost the price of a house. In just a few years, that was superseded by hardware that you could get for a couple of thousand dollars. Now it’s come down to a few hundred dollars, and we’re able to use it routinely. If not for the gaming industry, we wouldn’t have all of the graphics processor power that we need.

RALF KAEHLER: I completely agree with Stuart here. The ever-evolving capabilities of graphics hardware are very important for this work. You can now realize interactive visualizations of datasets that were far out of the reach of standard desktop workstations five or ten years ago.

TKF: With all of that computing power, how much of the process is science, and how much of it is art? If the three of you were to visualize the same event, would you end up with similar results?

RALF KAEHLER: I would say that there’s a lot of creativity involved in the process. It might be comparable to taking a photograph of some object. You have all of this freedom of how to choose your camera position, lighting conditions, color filters and so on. Similarly, with the same numerical simulation, you can end up with millions of different images by changing around these variables. So it really depends on the audience you’re targeting, what features in the original dataset you want to highlight, and what story you want to get across. If different people work on visualizations for the same dataset, the results can be totally different.

One of our most recent visualizations was a collaboration with the Hayden Planetarium at the American Museum of Natural History in New York. It shows the role of dark matter in forming the larger structures in the universe. For this audience, we took the term dark matter a bit more literally than usual. If we had done this rendering for scientists, we would have represented higher dark matter densities in brighter colors. But studies have shown that often confuses the general public. So in this visualization, we actually turned it around and rendered the dark matter in darker colors and added some background light. That helped guide the audience and clarified what was dark matter and what was not.

STUART LEVY: I agree. I think we should look at visualization like mapmakers look at map making. A good mapmaker will be deliberate in what gets included in the map, but also in what gets left out. Visualizers think about their audience, as Ralf says, and the specific story they want to tell. And so even with the same audience in mind, you might set up the visualization very differently to tell different stories. For example, for one story you might want to show only what it’s possible for the human eye to see, and in others you might want to show the presence of something that wouldn’t be visible in any sort of radiation at all. That can help to get a point across.

TKF: It sounds like theres quite a bit of room for artistic choice. For outreach purposes, then, why is it important for the visualizations to be based on scientifically accurate data? Why are you creating them, rather than a movie house?

RALF KAEHLER: Using sophisticated numerical simulations ensures that the science is depicted correctly. Besides this, it’s hard to model a lot of the phenomena in astrophysics using the artistic tools that Hollywood movies employ. The phenomena are just too complex to draw by hand. I think more and more of these artistic tools are now starting to incorporate some sort of simplified simulation codes in order to model things like explosions, to make it look more realistic.

TKF: When you’ve created visualizations for a public outlet, have you ever sat in the audience and watched the public’s reaction? What’s that like?

DYLAN NELSON: It’s kind of amazing, to be honest, the amount of press and public interest that’s come out of the Illustris project. Actually, just yesterday I got a call from my father, who had been browsing the news on his phone and saw an image from Illustris on the front page of The New York Times website. This was a still image that we made just for the purposes of putting it on the website, and it’s probably appeared in a dozen newspapers so far. It’s great that there’s so much interest, and that the images are becoming almost iconic.

STUART LEVY: For me, it’s great to watch visualizations in planetarium domes. It’s the most wonderful thing to lie down in the middle of a planetarium – or even in an IMAX theater – and just look up. Having the audience completely surrounded by what they’re seeing can be really breathtaking.

RALF KAEHLER: I love it when visualizations are shown in planetariums, too. It just looks so impressive – much more impressive than looking at the visualizations on a flat monitor in my office. I’ve worked on visualizations that were shown in places like the American Museum of Natural History and the Morrison Planetarium at the California Academy of Sciences. These are great places to reach a lot of people in a nice, inspiring environment. Even though when I’m sitting in the audience it’s too dark to gauge other people’s reactions, sometimes we get emails from people who saw the planetarium shows and write how much they liked it. It’s really motivating, and shows that our time is being well invested.

Subscribe to HPCwire's Weekly Update!

Be the most informed person in the room! Stay ahead of the tech trends with industy updates delivered to you every week!

Advancing Modular Supercomputing with DEEP and DEEP-ER Architectures

February 24, 2017

Knowing that the jump to exascale will require novel architectural approaches capable of delivering dramatic efficiency and performance gains, researchers around the world are hard at work on next-generation HPC systems. Read more…

By Sean Thielen

Weekly Twitter Roundup (Feb. 23, 2017)

February 23, 2017

Here at HPCwire, we aim to keep the HPC community apprised of the most relevant and interesting news items that get tweeted throughout the week. Read more…

By Thomas Ayres

HPE Server Shows Low Latency on STAC-N1 Test

February 22, 2017

The performance of trade and match servers can be a critical differentiator for financial trading houses. Read more…

By John Russell

HPC Financial Update (Feb. 2017)

February 22, 2017

In this recurring feature, we’ll provide you with financial highlights from companies in the HPC industry. Check back in regularly for an updated list with the most pertinent fiscal information. Read more…

By Thomas Ayres

HPE Extreme Performance Solutions

O&G Companies Create Value with High Performance Remote Visualization

Today’s oil and gas (O&G) companies are striving to process datasets that have become not only tremendously large, but extremely complex. And the larger that data becomes, the harder it is to move and analyze it – particularly with a workforce that could be distributed between drilling sites, offshore rigs, and remote offices. Read more…

Rethinking HPC Platforms for ‘Second Gen’ Applications

February 22, 2017

Just what constitutes HPC and how best to support it is a keen topic currently. Read more…

By John Russell

HPC Technique Propels Deep Learning at Scale

February 21, 2017

Researchers from Baidu’s Silicon Valley AI Lab (SVAIL) have adapted a well-known HPC communication technique to boost the speed and scale of their neural network training and now they are sharing their implementation with the larger deep learning community. Read more…

By Tiffany Trader

IDC: Will the Real Exascale Race Please Stand Up?

February 21, 2017

So the exascale race is on. And lots of organizations are in the pack. Government announcements from the US, China, India, Japan, and the EU indicate that they are working hard to make it happen – some sooner, some later. Read more…

By Bob Sorensen, IDC

ExxonMobil, NCSA, Cray Scale Reservoir Simulation to 700,000+ Processors

February 17, 2017

In a scaling breakthrough for oil and gas discovery, ExxonMobil geoscientists report they have harnessed the power of 717,000 processors – the equivalent of 22,000 32-processor computers – to run complex oil and gas reservoir simulation models. Read more…

By Doug Black

Advancing Modular Supercomputing with DEEP and DEEP-ER Architectures

February 24, 2017

Knowing that the jump to exascale will require novel architectural approaches capable of delivering dramatic efficiency and performance gains, researchers around the world are hard at work on next-generation HPC systems. Read more…

By Sean Thielen

HPC Technique Propels Deep Learning at Scale

February 21, 2017

Researchers from Baidu’s Silicon Valley AI Lab (SVAIL) have adapted a well-known HPC communication technique to boost the speed and scale of their neural network training and now they are sharing their implementation with the larger deep learning community. Read more…

By Tiffany Trader

IDC: Will the Real Exascale Race Please Stand Up?

February 21, 2017

So the exascale race is on. And lots of organizations are in the pack. Government announcements from the US, China, India, Japan, and the EU indicate that they are working hard to make it happen – some sooner, some later. Read more…

By Bob Sorensen, IDC

TSUBAME3.0 Points to Future HPE Pascal-NVLink-OPA Server

February 17, 2017

Since our initial coverage of the TSUBAME3.0 supercomputer yesterday, more details have come to light on this innovative project. Of particular interest is a new board design for NVLink-equipped Pascal P100 GPUs that will create another entrant to the space currently occupied by Nvidia's DGX-1 system, IBM's "Minsky" platform and the Supermicro SuperServer (1028GQ-TXR). Read more…

By Tiffany Trader

Tokyo Tech’s TSUBAME3.0 Will Be First HPE-SGI Super

February 16, 2017

In a press event Friday afternoon local time in Japan, Tokyo Institute of Technology (Tokyo Tech) announced its plans for the TSUBAME3.0 supercomputer, which will be Japan’s “fastest AI supercomputer,” Read more…

By Tiffany Trader

Drug Developers Use Google Cloud HPC in the Fight Against ALS

February 16, 2017

Within the haystack of a lethal disease such as ALS (amyotrophic lateral sclerosis / Lou Gehrig’s Disease) there exists, somewhere, the needle that will pierce this therapy-resistant affliction. Read more…

By Doug Black

Azure Edges AWS in Linpack Benchmark Study

February 15, 2017

The “when will clouds be ready for HPC” question has ebbed and flowed for years. Read more…

By John Russell

Is Liquid Cooling Ready to Go Mainstream?

February 13, 2017

Lost in the frenzy of SC16 was a substantial rise in the number of vendors showing server oriented liquid cooling technologies. Three decades ago liquid cooling was pretty much the exclusive realm of the Cray-2 and IBM mainframe class products. That’s changing. We are now seeing an emergence of x86 class server products with exotic plumbing technology ranging from Direct-to-Chip to servers and storage completely immersed in a dielectric fluid. Read more…

By Steve Campbell

For IBM/OpenPOWER: Success in 2017 = (Volume) Sales

January 11, 2017

To a large degree IBM and the OpenPOWER Foundation have done what they said they would – assembling a substantial and growing ecosystem and bringing Power-based products to market, all in about three years. Read more…

By John Russell

US, China Vie for Supercomputing Supremacy

November 14, 2016

The 48th edition of the TOP500 list is fresh off the presses and while there is no new number one system, as previously teased by China, there are a number of notable entrants from the US and around the world and significant trends to report on. Read more…

By Tiffany Trader

Lighting up Aurora: Behind the Scenes at the Creation of the DOE’s Upcoming 200 Petaflops Supercomputer

December 1, 2016

In April 2015, U.S. Department of Energy Undersecretary Franklin Orr announced that Intel would be the prime contractor for Aurora: Read more…

By Jan Rowell

D-Wave SC16 Update: What’s Bo Ewald Saying These Days

November 18, 2016

Tucked in a back section of the SC16 exhibit hall, quantum computing pioneer D-Wave has been talking up its new 2000-qubit processor announced in September. Forget for a moment the criticism sometimes aimed at D-Wave. This small Canadian company has sold several machines including, for example, ones to Lockheed and NASA, and has worked with Google on mapping machine learning problems to quantum computing. In July Los Alamos National Laboratory took possession of a 1000-quibit D-Wave 2X system that LANL ordered a year ago around the time of SC15. Read more…

By John Russell

Enlisting Deep Learning in the War on Cancer

December 7, 2016

Sometime in Q2 2017 the first ‘results’ of the Joint Design of Advanced Computing Solutions for Cancer (JDACS4C) will become publicly available according to Rick Stevens. He leads one of three JDACS4C pilot projects pressing deep learning (DL) into service in the War on Cancer. Read more…

By John Russell

IBM Wants to be “Red Hat” of Deep Learning

January 26, 2017

IBM today announced the addition of TensorFlow and Chainer deep learning frameworks to its PowerAI suite of deep learning tools, which already includes popular offerings such as Caffe, Theano, and Torch. Read more…

By John Russell

HPC Startup Advances Auto-Parallelization’s Promise

January 23, 2017

The shift from single core to multicore hardware has made finding parallelism in codes more important than ever, but that hasn’t made the task of parallel programming any easier. Read more…

By Tiffany Trader

Tokyo Tech’s TSUBAME3.0 Will Be First HPE-SGI Super

February 16, 2017

In a press event Friday afternoon local time in Japan, Tokyo Institute of Technology (Tokyo Tech) announced its plans for the TSUBAME3.0 supercomputer, which will be Japan’s “fastest AI supercomputer,” Read more…

By Tiffany Trader

Leading Solution Providers

CPU Benchmarking: Haswell Versus POWER8

June 2, 2015

With OpenPOWER activity ramping up and IBM’s prominent role in the upcoming DOE machines Summit and Sierra, it’s a good time to look at how the IBM POWER CPU stacks up against the x86 Xeon Haswell CPU from Intel. Read more…

By Tiffany Trader

Nvidia Sees Bright Future for AI Supercomputing

November 23, 2016

Graphics chipmaker Nvidia made a strong showing at SC16 in Salt Lake City last week. Read more…

By Tiffany Trader

BioTeam’s Berman Charts 2017 HPC Trends in Life Sciences

January 4, 2017

Twenty years ago high performance computing was nearly absent from life sciences. Today it’s used throughout life sciences and biomedical research. Genomics and the data deluge from modern lab instruments are the main drivers, but so is the longer-term desire to perform predictive simulation in support of Precision Medicine (PM). There’s even a specialized life sciences supercomputer, ‘Anton’ from D.E. Shaw Research, and the Pittsburgh Supercomputing Center is standing up its second Anton 2 and actively soliciting project proposals. There’s a lot going on. Read more…

By John Russell

TSUBAME3.0 Points to Future HPE Pascal-NVLink-OPA Server

February 17, 2017

Since our initial coverage of the TSUBAME3.0 supercomputer yesterday, more details have come to light on this innovative project. Of particular interest is a new board design for NVLink-equipped Pascal P100 GPUs that will create another entrant to the space currently occupied by Nvidia's DGX-1 system, IBM's "Minsky" platform and the Supermicro SuperServer (1028GQ-TXR). Read more…

By Tiffany Trader

IDG to Be Bought by Chinese Investors; IDC to Spin Out HPC Group

January 19, 2017

US-based publishing and investment firm International Data Group, Inc. (IDG) will be acquired by a pair of Chinese investors, China Oceanwide Holdings Group Co., Ltd. Read more…

By Tiffany Trader

Dell Knights Landing Machine Sets New STAC Records

November 2, 2016

The Securities Technology Analysis Center, commonly known as STAC, has released a new report characterizing the performance of the Knight Landing-based Dell PowerEdge C6320p server on the STAC-A2 benchmarking suite, widely used by the financial services industry to test and evaluate computing platforms. The Dell machine has set new records for both the baseline Greeks benchmark and the large Greeks benchmark. Read more…

By Tiffany Trader

Is Liquid Cooling Ready to Go Mainstream?

February 13, 2017

Lost in the frenzy of SC16 was a substantial rise in the number of vendors showing server oriented liquid cooling technologies. Three decades ago liquid cooling was pretty much the exclusive realm of the Cray-2 and IBM mainframe class products. That’s changing. We are now seeing an emergence of x86 class server products with exotic plumbing technology ranging from Direct-to-Chip to servers and storage completely immersed in a dielectric fluid. Read more…

By Steve Campbell

What Knights Landing Is Not

June 18, 2016

As we get ready to launch the newest member of the Intel Xeon Phi family, code named Knights Landing, it is natural that there be some questions and potentially some confusion. Read more…

By James Reinders, Intel

  • arrow
  • Click Here for More Headlines
  • arrow
Share This