In this blog, Marcus Bond, an XSEDE Campus Champion and professor of inorganic chemistry and x-ray crystallography at Southeast Missouri State University, shares his experience at the ISC 2018 Sunday program.
Knowledge. As a relative newcomer to the field of high performance computing (HPC), this is primarily why I attend. The last few years have been a flurry of HPC conferences, workshops, and Linux Cluster Institutes as I try to educate myself and become more conversant about supercomputing. Repetition is helpful, since it is easy to forget from one conference to the next, but so are the different perspectives the venues bring, which is a major reason why I am here for my first ISC.
Tutorial Day, of course, is all about learning. As a newcomer, any tutorial with “Getting started,” “Beginner’s,” “Basic,” or “Introduction” immediately catches my eye. Hence, choosing the morning tutorial “Getting started with Containers on HPC through Singularity” was an easy choice. For me and about 100 other people because that tutorial was packed…a fact that the presenter, Carlos Eduardo Arango, noted with some amazement.
I am interested in implementing containers on our small system. We have reached a point where users are starting to prototype applications on their desktops using a different flavor of Linux, and I’m then stuck trying to figure out how to install it on the cluster. I have been to talks on containers before, but I would have to say that this was the most lucid and compelling explanation I have heard. There appears not to be much of a performance hit, and perhaps a bit of a performance boost, according to data in the first presentation, since users can employ libraries optimized for their applications rather than those installed on the bare metal. On the cluster, users run in their own account, using the batch scheduler. It seems like a perfect solution, and I can’t wait to get home and try it.
Lunch allowed some time to get a bird’s eye view of the ISC Exhibition Floor setup. What will be even more amazing is to watch how fast it all comes down the minute it closes!
For the afternoon tutorial I attended “Better Scientific Software.” I do not write software, at least not at the moment. However, as an X-ray crystallographer, I am a longtime user of scientific software since the crystallography field has been involved with computing almost since the advent of electronic computers. (Crystallographer David Sayre, for example, was on the original team that wrote the first Fortran compiler). Therefore, I’m interested in the process of writing scientific software…especially better scientific software.
The tutorial started with a description of “heroic programming,” in which a single programmer puts in the extra effort to finish a project, and the resulting greater cost in technical debt that increases the future work required to sustain the software. Crystallographic software has suffered the parallel problem of single individuals who have developed and maintained the mainline programs…and nobody else really knows the code. With these individuals now in advanced age, the crystallographic community has made a decided effort to develop new software packages in a more sustainable way.
The central part of the tutorial surprisingly dealt with the soft skills of people and project management. As a participant in the University of Oklahoma-sponsored Advanced Cyberinfrastructure Research and Facilitators Virtual Residency, which does focus on developing soft-skills, I found this part of the tutorial particularly useful. I was unfamiliar with the Kanban method for tracking workflow, and most critically limiting the number of in-progress tasks. While this was presented in the context of software development, it well applies to other projects, and was my biggest takeaway.
Oh, and I was happy to see crystallography rear its (in this case ugly) head:
A full accounting of Marcus’ ISC18 experience can be read on the STEM-Trek blog site.