ISC Student Cluster Competition: TeamEPCC's training session with sponsor Boston Ltd

Author: Xu Guo
Posted: 13 Jun 2019 | 10:00

TeamEPCC will participate in the ISC19 Student Cluster Competition on 17–19 June in Frankfurt, Germany. Please come to visit us at Booth A-1452

TeamEPCC undertook a special two-day training in May provided by our team’s sponsor Boston Ltd to further prepare for the ISC19 Student Cluster Competition (SCC). During the training, the team had plenty of opportunities to assemble the hardware and experiment with multiple system softwares.

Prior to the training

Before to the training at Boston Ltd., we tested multiple configurations and setups for MPI, CUDA-aware MPI, Remote Direct Memory Access (RDMA) and compilers, etc, based on the test nodes provided by Boston. We also tried to figure out the different requirements for installing and optimising the competiton's announced benchmarks and applications.  

Travelling to Boston Ltd

We all got up at 4am to travel from Edinburgh to St Albans, arriving at Boston's headquarters just in time! We were warmly welcome by the Boston staff and introduced to the facilities. The campus includes offices for around 40 employees, a lab for the engineers, and a big warehouse for the retail material including servers, racks and pre-configured solutions, etc.

Building the cluster

Patrick (Paddy) O’Neill, the HPC system engineer who has been the main contact from Boston supporting us for ISC19 SCC, introduced the hardware to us and also allowed us to have some hands-on experiences! We looked into different components of the cluster, tried to unpack them first, then put them back together, plugged them and assembled the individual nodes, etc. Besides the Intel Skylake CPU nodes, we also tried to assemble some of the NVIDIA Tesla V100 GPU nodes in the temporary Nvidia DGX station which was built for training purposes. We then used Warewulf Cluster Manager to combine individual nodes into a cluster. By following the installation guides and with assistance from Paddy, we eventually had a working cluster!

After the hardware assembly, with the guidance from Paddy, we wiped the main memory, as a new file system was going to be set. We also learned about how to establish an automatic replication system, remote BIOS access and power monitoring, system tuning, etc.

Meeting HPC engineers at Boston Ltd

During the 2-day training, we had some opportunities to speak to HPC engineers working for Boston Ltd. Although usually based at the headquarters in St Albans, they also need to travel a lot to support customers in person. Boston Ltd has just opened new offices in South Africa, Germany, India and Australia, and we were told that they are hiring now! We have no doubt working at Boston Ltd would be very exciting.

Our thanks

We all benefited a lot from this special experience of assembling, installing and tuning our competition cluster by ourselves. Many thanks to Boston Ltd for providing not only the competition cluster and more hands-on experience of it, but also such a precious opportunity for us to get to know more about HPC industrial life. We’d also like to specially thank Paddy for all his help and great teaching!

TeamEPCC will participate in the ISC19 Student Cluster Competition on 17-19 June in Frankfurt, Germany. Please come to visit us at Booth A-1452. We will also share our photos and videos of our activities at ISC. See:
Instagram: teamepcc
Facebook: Teamepcc
Twitter: @teamepcc




Blog Archive