Image: Vsualization of the Universe
Argonne National Laboratory
This visualization of the Universe as it condenses around fluctuations in the density of dark and ordinary matter is a result from a collaboration between Argonne National Lab in Illinois and the San Diego Supercomputer Center in California.
updated 1/29/2010 12:53:57 PM ET 2010-01-29T17:53:57

Supercomputing has helped astrophysicists create massive models of the universe, but such simulations remain out of reach for many in the United States and around the world. That could all change after a successful test allowed scientists in Portland, Ore. to watch a Chicago-based simulation of how ordinary matter and mysterious dark matter evolved in the early universe.

The streaming event also took place in real-time, which means that teams in both Chicago and Portland could have theoretically played together in the simulation as easily as PC or console video gamers play together in online games.

The demo goes far beyond entertaining people with pretty 3-D journeys through the early universe. Only supercomputers can handle the huge amounts of data that make up the most sophisticated astrophysics models, and scientists can't always travel to places with supercomputing clusters to do their research. Having the ability to stream a fully-rendered simulation online allows scientists to collaborate on research remotely and overcome the barriers of limited access to supercomputers.

"This is an example of trying to break down that barrier — a barrier that gets higher every day as simulations get more complex," said Mark Hereld, a computer scientist at Argonne National Laboratory (ANL) in Illinois.

Complex simulations have become necessary for tackling the tougher astrophysics puzzles such as dark matter. Scientists estimate that dark matter makes up over 70 percent of mass in the universe, but they can only detect the invisible substance by measuring the gravitational effect it has on visible ordinary matter. The latest simulation showed how ordinary matter and dark matter might interact over the course of almost 7 billion years, starting from the theoretical Big Bang at the beginning of the known universe. The universe is 13.7 billion years old.

Cracking an astrophysics puzzle
Researchers specifically used the simulation to gauge how well they could detect Baryon Acoustic Oscillations, a phenomenon related to the clustering of certain fundamental particles in the gas between galaxies. They hope that the simulation will allow upcoming astrophysics surveys to make a direct comparison with the supercomputer predictions.

"We can measure the density of intergalactic gas by seeing how strongly it absorbs the light from distant objects, particularly quasars," said Rick Wagner, an astrophysicist at the University of California in San Diego. He told that Baryon Acoustic Oscillations show up as "peaks" in the overall density map of the universe, and added that "measuring these peaks accurately is one of the best tools for nailing down the fundamental properties of the universe."

Such subtle density differences only show up on huge galactic scales, and so the model simulated a volume of space 1 billion light-years on each side. One light-year represents the distance that light can travel over the course of a year – about 6 trillion miles (10 trillion km).

Supercomputing power
Turning such huge chunks of data into an interactive movie of the universe proved tricky. The simulation began as 148 terabytes of data, where one terabyte roughly equals 200,000 digital photos or music mp3s. The University of Tennessee's Kraken supercomputer spent the equivalent of 4 million CPU hours crunching the numbers, before it could pass on the data to the Eureka computer at ANL. Eureka then took care of the visual rendering that transformed the simulation into a 3-D model. Slideshow: Month in Space: January 2014

The true test came when ANL shared its rendered simulation in real-time on a massive tiled display hosted by the San Diego Supercomputer Center at the 2009 Supercomputing Conference held in Portland, Oregon during last November. Hereld's group developed software to stream the simulation over a fiber optic connection at 10 gigabits per second, or about 10,000 times faster than the average cable modem speed for U.S. broadband users.

Wagner spent much of his time in Portland managing the tiled display wall that allowed him and his colleagues to view the simulation on a huge scale, and identify subtle patterns. The display walls also provide enough visual space to show many different activities at the same time, ranging from video conferencing to movies and presentations.

"This wall has personally given me a more intuitive understanding of how the structures (e.g., galaxies, galaxy clusters, filaments) at various scales relate to one another," Wagner said.

Simulations for all
Back in Chicago, Hereld said that it's just a matter of tweaking the software coding to allow remote end-users such as Wagner to interact with data or simulations. Researchers can already take pre-rendered chunks of simulations home to play with on their laptops or desktop computers, but that still pales in comparison to being able to dive into a full-fledged simulation at will — even if they are physically hundreds or thousands of miles away.

"If your science is the discovery science where you have to interact, hop around and look for stuff in a simulation, that's the time when you need to do the rendering in real-time," Hereld said.

Having real-time access to the most powerful simulations and datasets could ultimately prove a game-changer for the greater scientific community.

"Our data came from a supercomputer, but it could have come from a large scale astronomical survey, like Pan-STARRS or LSST — or a particle collider," Wagner said. "The freedom to move our data allowed us to see it in ways we never thought possible; it will be great to see this capability made available to more researchers."

© 2013 All rights reserved. More from


Discussion comments


Most active discussions

  1. votes comments
  2. votes comments
  3. votes comments
  4. votes comments