Super Computing 2016


In November 2016 members of the ACI-REF consortium had the opportunity to attend Super Computing 2016 (SC16). Below are some of their experiences.

University of Wisconsin, Madison – Lauren Michael

For Wisconsin, this was our very first time having an official campus presence, and we would say that it was a success! Not only did we make a lot of fruitful contacts with other campuses around the ACI-REF and CaRC projects, but we also participated in conversations that included a number of our academic and industry partners who have a well-established contribution to the SC16 culture.

 

Our shared booth with Clemson University was a focal point of our SC16 presence, including shared space for engagement and in-booth talks. For our own talks (three that we gave on a repeating daily schedule), we focused on our unique campus approach to supporting research computing (flagship talk talk), which includes the personalized support our Research Computing Facilitators provide (second talk), as well as local expertise in high-throughput computing within the Center for High Throughput Computing, including our production of the HTCondor compute scheduling software (our third talk). As these were the major aspects of the Wisconsin computing brand that we hoped to communicate, we were more than satisfied with the quality and quantity of engaging conversations on these very topics, especially with representatives from other academic institutions. In addition, we also had an HTCondor demo in the Google booth where HTCondor’s Jamie Frey presented the ability to dynamically expand a local HTCondor cluster into the cloud. With help from our facilitators, Jamie used a real research example from our campus’s Small Molecule Facility as the computational fodder for Google’s cloud services to chew on. Our team also included Jan Cheetham from the CIOs office, Todd Miller who presented on HTCondor in our booth, and Chris Harrison, who collaborates with CHTC in running compute and data systems for the Department of Biostatistics and Medical Informatics in the Medical School

 

Lead ACI-REF Lauren Michael collaborated with Marcin Ziolkowski of Clemson to share “booth boss” responsibilities and also worked together on in-booth presentations for the ACI-REF and CaRC projects, and joint engagement by both schools was another important success! As we move into Phase II of ACI-REF and kick off the CaRC project, we believe we were able to get a lot of campuses at SC16 interested in joining CaRC and hiring facilitators in the ACI-REF model.

Clemson University – Marcin Ziolkowski

Clemson University has been present at Supercomputing conference for many years focusing on promotion of Clemson University as a major High Performance Computing player among public universities in the United States. SC16 was a great opportunity for faculty and students from Clemson to present their research on applications of HPC platforms in big data research, genomics, molecular modeling, software development, earthquake simulations, cloud computing research and low power software design. This year Clemson has partnered with University of  Wisconsin, Madison on presenting joint projects including ACI-REF, CaRC and CloudLab. The presence at SC exhibit floor gives Clemson great opportunity to engage with the research computing community and share experiences about the platforms we are using, humanware in research computing and projects done using technologies presented at the exhibit floor. SC is a place where Clemson can attract students interested in graduate programs and start new collaborations. Among many aspects of SC, the networking with other universities is a very important reason to be there.

University of Utah

 

Wim:

It was quite a few years since I last attended SC. A few technologies sparked my interest.

Among them, the Kokkos library (part of Trilinos) which allows users to develop C++ code that can run on CPU, GPU and Intel Phi Architectures. The Kokkos developers are hoping that the their library becomes part of the new (upcoming) C++ standard.

Intel has made significant contributions/investments to speed up the numerical python libraries (NumPy, SciPy). Intel now provides a Python Suite (Intel Python) which has been optimized for the Intel processors). Besides their investments in Python they are also paying attention to other scripting languages such as Julia, R.

In this version of SC the GPUs & deep learning were playing a central role.

It was refreshing to be back at SC.

 

Martin:

I was glad to be back at SC after a multi-year hiatus. The conference’s size make deciding what to attend and whom to meet a complicated optimization exercise. What I enjoy the most is getting updates on both technology and programming approaches/standards. This SC this included the ongoing struggle to come up with a viable standard for multi-core and accelerator programming. The several meetings I attended at SC are a good direction to tackle this problem but the solution is still out in the future. SC also provided a good opportunity to meet people that I have only interacted with online before.

 

Anita:

While CHPC has had an SC presence for many years, this was the first time I attended the conference.  I enjoyed being able to participate in the booth – from the setup, to talking to booth visitors about the University of Utah and CHPC, to the tear down — instead of only hearing about the process.  It was also interesting to visit the booths of other research computing centers and to see what they were highlighting about their centers and institutions.

 

A second highlight was the workshops and presentations focused on HPC User Support Tools and on Training & Education offerings which I attended – basically hearing about what other centers were doing to support their user base and to find out about new directions that we can explore.  I came back to CHPC with a couple packages I would like to try in order to improve both our ability to troubleshoot user issues (SanityTool from TACC) as well as to better provide information on the use of our applications (getexample from Michigan State University).

 

University of Southern California – Erin Shaw

I attended SC16 as one of eight staff members from USC’s Center for High-Performance Computing (HPC), which has had an Exhibitor booth at SC for the past 13 years. The booth has two tables and a small presentation area where USC researchers give hourly presentations. HPC staff work a few hours a day answering questions and assisting speakers. It was my second year at SC16 and this year my goal was to learn about GPU programming and seek out training opportunities, especially with MATLAB, INTEL and NVIDIA. The bigger booths had some excellent presentations. I attended the HPC facilitation and education workshops again this year, as well as a BoF on K12 outreach that was co-presented by one of our SCEC researchers. Most enjoyable were the HPC team dinners, meeting up with our ACI-REF colleagues, and what I can only hope is the annual craziness of the reception party, hopefully with a better band next year!