SGCI Client Chem Compute Meets Goals
By Nayiri Mullinix
“I was stuck using a small server provided by my University but, by working with SGCI, I was able to transition to much better hosting on Jetstream. Now our Chem Compute science gateway can serve many more undergraduate students simultaneously, and classes are already taking advantage of that this Fall.
Prior to SGCI support we were submitting jobs only using a local server and Comet and could only work on 8 jobs at once. The SGCI Extended Developer Support consultant I worked with, Sudhakar, suggested we apply for supplements on Bridges and Jetstream to be able to support more job submissions as well as long jobs. We quickly set up on Bridges, which helped with reliability, and then setup Jetstream as a virtual private elastic cluster, which dramatically reduced queue times (from minutes to seconds), which was important for our undergraduate audience. We’ve since had about 6,000 jobs submitted to Jetstream and 2,700 on Bridges, which wouldn’t have been possible without the support from SGCI.”
The Chem Compute science gateway had humble beginnings as a small classroom operation. Prior to engaging with SGCI, Mark Perri, Associate Professor of Chemistry at Sonoma State University, had created a system that would give his undergraduate students compute resources through a server that was in his lab. The server would accept submissions and data but, as it was a fairly low-end desktop system, it would crash when too many users logged in. This problem intensified with his interest in adding Chem Compute to ChemWiki. This free, online alternative to textbooks would require him to be able to host many more logins than was possible on his classroom system and made him realize he would need to develop a renewed version of Chem Compute. In doing so, not only would he be bringing down barriers to computational chemistry resources at Sonoma State and accelerating student access to important tools, but he’d also be able to make the resource accessible to anyone from anywhere. Perri also recognized the importance of integrating built-in experiments and on-screen instructions to make sure it was an easy tool for both students and faculty to use.
Perri started his quest to grow Chem Compute by submitting an application to XSEDE, thinking that it would help him immensely to be able to use the Comet supercomputer for calculations. SGCI’s Marlon Pierce, who leads the Extended Developer Support service area, was on the review for Perri’s XSEDE application and quickly recognized that his needs would be well-suited for an engagement with SGCI. As Perri’s aim was to support more users and more job submissions, the advice, guidance, and technical help he’d receive from SGCI could help him achieve his goals. After learning what SGCI had to offer, Perri said, “One major reason why I wanted to work with SGCI is that I’m just one person working on this, and I can only do one thing at a time. I’ve got to get website capabilities, users, and content, and bouncing between those three things as an individual can be overwhelming.”
By engaging with SGCI, Perri was, indeed, able to achieve his goals. Chem Compute was set up on the Django (1.10) Python 3 web framework and also uses Celery, RabbitMQ, and Mariadb. The gateway now uses Jetstream, an SGCI partner, for hosting the portal and an elastic virtual cluster for processing computations. When reflecting on the work he did with SGCI, Perri says that the process went smoothly and all the steps of development were clear from the start when a work plan was established. Most of the work was executed with the help of SGCI’s Sudhakar Pamidighantam, who was able to offer not only his expertise but also the expertise of other Extended Developer Support team members.
The new Chem Compute is now up and running, and enjoying a growing user base. People from all around the country have been discovering and using the gateway, making it more challenging for Perri to keep track of who his users are. Perri has made positive observations from his class usage, however, and is pleased with the fact that his students are able to complete experiments using Chem Compute and that they all feel comfortable and significantly more confident solving chemical problems using the new and improved web platform. Just last week, Perri’s students submitted 2,249 jobs on Jetstream.
Asked what advice he would offer to anyone considering requesting SGCI services, Perri said, “I would definitely tell them to pursue services with SGCI. It’s a big benefit to talk with other people who do the same things and to not work alone because you can waste a lot of time by working alone.”
What’s next for Chem Compute? Perri is continuing to work on recruiting new users, and he’s pretty sure he knows how he’ll engage with SGCI to achieve his next goal: improved usability and user-centered design.
Figure 1. Chem Compute homepage, offering two packages: GAMESS (Electronic Structure Calculation) and TINKER (Molecular Dynamics)
Figure 2. GAMESS output is visualized using JSmol, and spectra can be calculated and displayed as well.
Figure 3. Tinker trajectory output is visualized using ChemDoodleWeb.
Web-Based Job Submission Interface for the GAMESS Computational Chemistry Program