ISC 2016 Student Cluster Competition at Germany and First Visualization Tutorial
I spend most of the third week at the ISC 2016 Student Cluster Competition at Frankfurt, Germany. My mentor from my last internship at Lawrence Berkeley National Laboratory (LBNL) connected me and 5 more girls that had also interned at LBNL last year with Rebecca Hartman-Baker, who has experience mentoring previous groups for Student Cluster Competitions. She wanted to make an all girls team to compete at this year's ISC Student Cluster Competition, so she made the team with us and we meet virtually every week on last Spring's semester.
From left to right, Elizabeth Wang (Monta Vista HS), Ruoyun Zheng (Monta Vista HS), Kristi Arroyo (Missouri U of S&T), Yong Li Dich (Emily) (Harvard U),Davina Ngai (Queen's U), and Grace M. Rodriguez (U of Puerto Rico)
The competition was intense. They were 12 teams competing from all over the globe. We arrived at the conference center on Saturday to set up our cluster. On Sunday, we started running tests for our applications to make sure Ada (we name our cluster Ada in honor of Ada Lovelace) was working well. We didn't encounter major problems running tests with the applications, so we left to eat later on that day happy. So far, so good.
The competition started on Monday at 3:30 GMT. The first challenge was a T-shirt scavenger hunt, where we have to look for our shirts all over the conference center. Only when the team members had all 6 t-shirts could they get the usb and begin running the applications.
The challenge of the day was to run the application HPCC benchmark, which measures the floating point rate of execution for solving a linear system of equations. We were able to run it successfully.
The second day we started at 8:00 am GMT. We had three applications to run, including mine, which is Splotch, an open ray-tracing software. Running all the applications proved to be more challenging. The input I was given to run my application created 2000 images. I made the mistake of running the mpi command with the given input incorrectly, so it started to run in the head node, and it took one hour and half to create 200 images. So I was only able to hand in about 15% of the required work. We also had some trouble with the other application, WRF, who prove very difficult to compile. Ruoyun was able to run her application successfully (Graph500), so at least we were able to complete running this app. Overall, it was mentally challenging, but I feel like we learned a lot. There was a moment when we had employees from Intel, Cray and the Pawsey Supercomputing Center explaining us how to efficiently run in parallel (this was allowed in the second day as long as they didn't type anything directly on the head node).
The third day we had two mystery apps, but which was actually another input for WRF and an application named Clover-Leaf. The challenge was to run this applications using as less power as possible, so we had until the afternoon to test them and make the proper changes to the code and the cluster. We were able to test the apps efficiently. However at 12:00pm our nodes went down. We tried as much as we could to fix them; we even had two of the ISC committee try to help us. When were supposed to run the applications, we only had 3 nodes available. We then try to run Clover-Leaf with mpirun, but it wasn't recognizing the command. We then discover that the head node had some kind of problem, so we couldn't even do mpirun on the head run, and Cloverleaf could only be run using parallel. Therefor, we couldn't run any of the commands, so we rebooted the entire cluster, which mean we were disqualified from this day's challenge.
Overall, Shanghai Jiao Tong University from China won 3rd place, Tsinghua University from China won 2nd place and Centre for High Performance Computing from South Africa won 1st place. Huazhong University of Science and Technology from China won the Linpack award and Universitat Politècnica De Catalunya Barcelona Tech (UPC) from Spain won the Fan-favorite award. They later told us we got second for fan-favorite. Almost there but not quite there yet!
Even if we didn't won any awards, I feel so honored to have the opportunity to participate in this year's competition. I feel like I learned a lot, a lot of things that are not teach in my university. I also meet a lot of interesting people from all part of the world. Most people were very friendly and nice. Last but not least, I go to see the beautiful city of Frankfurt, eat German food (which was really good!) and try their chocolate. Definitely an experience to remember!
Meanwhile, I was able to come back to TACC last Friday. We had our first Visualization training, which was mostly a basic introduction to matplotlib, which is a python 2D plotting library that produces publication quality figures in a variety of hardcopy formats and interactive environments across platforms. It seems pretty easy to use, and has a lot of templates you can use for any graph you need to make. I wish I had knew about it sooner, would had been useful to use for my research.
Challenge of the week: Catching up to all of this week work and dealing with jet lag. Since I was busy with the competition almost the whole week, I am a little behind on watching tutorials and practicing Unity.
How I overcame this challenge: I'm planning on trying to work as much as I can on the weekend so I can catch up. I was able to call Marvis while at Frankfurt so she could gave me the updates on the process of our project. Apparently, the researcher who is the main source for the project contacted Andrew, and she is going to do the storyline. So we're waiting on her to give us the storyboard so we can begin developing the project in Unity. Dr. Brian also recommended we start filming in 360 video so we can have something for Xsede.
