TeamEPCC for ISC’18

Author: Guest blogger
Posted: 4 Jun 2018 | 20:41

Spyros Nita is an MSc student on our High Performance Computing with Data Science programme. He tells us about TeamEPCC, which will be competing in the Student Cluster Competition at ISC High Performance 2018.

Summer is on its way, and with it the annual International Supercomputing Conference (ISC). One of the three most important annual HPC events in the world, ISC focuses on the development of HPC, its scientific applications and its adoption into commercial environments. An important part of this event is the Student Cluster Competition (SCC), where teams representing twelve international universities meet to compete with each other. The University of Edinburgh is represented by TeamEPCC, four students currently enrolled in the MSc in High Performance Computing or MSc in High Performance Computing with Data Science.

The Student Cluster Competition will run for three days and take place from 25-27 June, in Frankfurt. With the help of our industry partner Boston Limited, we designed, built and set up a world-class cluster to compete in this year’s SCC. Boston Ltd has provided great support for the team and helped us reach award-winning scores since ISC’13. The goal will be to achieve the highest performance of the system within a power limit of 3000W. A series of benchmarks and applications are given, which every team member will use to carry out individual research.

Meet the team

Emmanouil (Manos) Farsarakis is our coach and mentor for the competition. He comes from a Greek-American background and has training in both Physics and high performance computing.

Being an Applications Consultant at EPCC, Manos participates in ground-breaking exascale and data science research as well as innovative HPC-education activities. Manos has been involved in the Student Cluster Competition since taking part as a student himself and contributing to the success of EPCC's “Highest Linpack” award-winning team at ISC’14 SCC.

Inaki Abadia is from Tudela, Spain. He holds a bachelor’s degree from University of Zaragoza, in Engineering Informatics. He specialises in computer architecture, embedded systems and data centres. He is currently enrolled in the MSc in High Performance Computing at the University of Edinburgh. His wide range of skills, from experience in Linux system administration and networking to a deep understanding of low-level computer architecture, make him a potent member of the Team. For his personal project, Inaki will focus on Grid application, a data parallel C++ mathematical object library. It defines its own basic data and vector types to internally exploit SIMD instructions depending on the underlying architecture. It also uses MPI to decompose tasks and OpenMP to vectorise loops.

Linta Koletsou from Greece did her undergraduate studies in Electrical and Computer Engineering at Aristotle University, Thessaloniki. She participated with team P.A.N.D.O.R.A., in the RoboCup Rescue competition, where they developed a robotic rescue vehicle and won second place in the autonomous class section. She is currently a postgraduate student undertaking the MSc in High Performance Computing at the University of Edinburgh. She has experience in numerical algorithms and parallel programming for GPGPUs using the CUDA API. For her personal project Linda will focus on the Nektar++ application, designed to support the construction of efficient partial differential equation (PDE) solvers using the spectral/hp element method.

Wilson Lisan from Indonesia graduated in the Electronics and Instrumentation Study Program at Universitas Gadjah Mada. He is currently a postgraduate student undertaking the MSc in High Performance Computing at the University of Edinburgh. His experience in computer hardware, such as building small beowulf clusters for real computing usage and overclocking (processor, GPU, RAM) on numerous architectures, make him a key member for the performance benchmarks taking place the first day of the competition. Wilson is interested on all three: HPL, HPCG, and HPCC benchmarks. He will carry on investigating HPCC benchmarks on different machines and compare it with HPL and HPCG in terms of performance, power consumption, and power efficiency.

Spyro Nita from Greece holds an undergraduate degree from the School of Informatics and Telecommunications at the University of Piraeus. During these years, Spyro specialised in Software Engineering and Intelligent Systems. He is currently a postgraduate student undertaking the MSc in High Performance Computing with Data Science at the University of Edinburgh. He has good knowledge of software engineering and object-oriented programming. His interest in cluster systems and data analytics with HPC make him the appropriate person for the AI application, which will involve training an image recognition model using deep convolutional neural networks. The goal will be to process the highest number of images per second while retaining a high prediction accuracy. The TensorFlow distributed training benchmark will be used based on the Imagnet dataset. Spyro will continue the project by comparing the deep learning performance on two different system architectures.

Visit us at Booth A-1453 at ISC

Even though the competition will keep us busy, we will always be delighted to show visitors our cluster and explain how we built it, its configuration and how the components complement each other to provide the highest possible performance and efficiency within the competition's power limit. We would love to interact with visitors to discuss various options and get useful feedback. We believe that our remarkable work and impressive results will pique the interest of all the attendees and our preparation for our booth visitors' experience will reward everyone who drops by.

Team EPCC has various social media accounts: Twitter (https://twitter.com/teamepcc), Facebook (https://www.facebook.com/TeamEPCC), and Instagram (https://www.instagram.com/teamepcc). Our team members take good care to keep them up to date and to reply to all our followers. During the conference week, we plan to share photos and videos of the activities at our booth through these accounts. Working on HPC is something we really love to do, and this competition motivates and inspires us to deliver high quality work.

EPCC's MSc programmes: www.epcc.ed.ac.uk/msc