User of the Month, May 2014: Robert Edwards
May 27, 2014 | Laurel Wamsley
We are happy to announce that the Globus user of the month for May 2014 is Robert Edwards, a staff scientist in the Theory Group at Jefferson Lab, a Department of Energy Nuclear Physics experimental facility.
Here he tells us about the work he’s doing, and how Globus helps him do it:
"We are nearing the end of a $310M upgrade of Jefferson Lab. There will be new experiments at the lab, in particular the GlueX experiment in Jefferson Lab’s new experimental Hall D. A major focus of the experiment is to search for the existence of 'exotic mesons' in the particle spectrum coming from QCD - the theory that describes how quarks and gluons are bound to make up the matter we know, like protons and neutrons. The existence of these exotic mesons indicates there are very non-trivial ways that the quarks and gluons can interact, and have never been clearly observed before this experiment.
In particular, I’m making 'lattice QCD' calculations of the spectrum - specifically looking for numerical evidence for the existence of such exotic states of matter. Our project is using the National Center for Supercomputing Applications (NCSA) Blue Waters computer facility - the largest of the NSF-supported supercomputers. The data I’ve transferred from Jefferson Lab to Blue Waters is to set up the calculation. The bulk of the transfers is data I have been transferring from Blue Waters to Jefferson Lab that are called 'quark propagators', and are just the solution to a large, sparse, linear system of equations. These propagators tell how a quark moves from one part of the space-time lattice to another part of the lattice. We tie these quark propagators together to effectively replicate how an exotic meson would move, or propagate, in space and time.
I’ve been using Globus for these transfers and have been absolutely delighted with its reliability and speed. Of course, both NCSA and Jefferson Lab have reasonably large data pipes, so that helps with transfer times, but Globus has been quite efficient at ramping up and down the transfers. In particular, I’ve transferred something like 113,000 files from NCSA back to Jefferson Lab, and this all went on with very few glitches - at least that I could noticeably observe. Namely, I do check the files on transfer back to Jefferson Lab. I found a very few number of failed transfers - empty in size on transfer; this is a fantastic level of reliability. I find it fair to say that trying to manage this kind of transfer in the 'bad ole days' via scripts wrapped over SCP would have been an absolute nightmare, and I would have been very hesitant to carry it out. Globus made it possible with a minimum of fuss - well up to the level that I had to initiate a large number of transfers for a huge number of files, but even then a good chunk of this was conveniently handled by transferring entire directories.
Many thanks for Globus. It really is 'enabling science'."
Thanks, Robert! We’re proud that Globus is part of the great research you’re doing.