Entire galaxies are moving around in the ‘stockyards’ of cluster galaxies

Understanding the history of galaxy and cluster formation fundamentally affects our knowledge of the Universe as a whole, and sets the context for our place within it, says Tom Quinn. An astronomy professor at the University of Washington, Quinn leads the N-body Shop, where he works on running and analyzing simulations of structure and planet formation in the Universe, along with studying galactic and solar system dynamics. He’s currently using the Blue Waters supercomputer at the University of Illinois’ National Center for Supercomputing Applications (NCSA) to model the formation and evolution of a population of galaxies in a coma-sized galaxy cluster, including their contribution to and interaction with the intra-cluster medium (ICM). He recently sat down with NCSA’s Barbara Jewett to discuss his work.

Tell me about your work.

What progress have you made using Blue Waters towards learning the answers to these questions?

There’s a huge dynamic range problem here. We want to have a high fidelity simulation of galaxies but we also want to do that for hundreds of galaxies within a cluster. So we essentially have cutting edge simulations of individual galaxies that we’re now putting in a bigger simulation of the cluster and doing that all at once, and that is just—you need state of the art computing resources to do that.

So is even a machine like Blue Waters big enough to do that?

Well, I’ve got a significant allocation, I’m running on 4,600 nodes—that’s, what 20% of Blue Waters?—to do that sort of simulation. And, I’ll say “ahhhh, I have not done the largest cluster yet.” (laughs) And that’s just one cluster. We really should be doing a population of clusters. If you’re asking me if I’m looking forward to the next generation machine, I’ll say very much so! (laughs)

I’ve been sort of bragging about how high resolution our galaxies are, but we can do better that way as well. It’s sort of the nature of astrophysics. The range of length scales of the universe, that’s the upper length, but, particularly, what drives what we can see? Well, that is star formation. And that happens at scales that are much below the size of an individual galaxy—you want to get into subparsec scales and galaxies themselves are a hundred kiloparsecs so that’s 10⁵ right there, and then you want to put that inside a megaparsecs sized cluster, and that’s another order of magnitude. And that’s sort of the basic scale problems you have with astrophysics that’s going to keep pushing computational limits for a while.

We’re talking here at the Charm++ workshop. I imagine that in order to do your research, the code is very important. That is, you have to be able to get the code to very efficiently create the clusters in these simulations.

In particular, clusters are a hard problem because most of the computation is concentrated in the center of the cluster. The Charm code’s adaptive run time system helps me distribute that concentrated load over a large number of compute nodes.

And maybe one day you’ll get your answer.

(laughs) No, I’ll get more questions!

Disclaimer: Due to changes in website systems, we've adjusted archived content to fit the present-day site and the articles will not appear in their original published format. Formatting, header information, photographs and other illustrations are not available in archived articles.