EPCC (Edinburgh Parallel Computing Centre)

I was a graduate student at the Edinburgh Parallel Computing Centre (EPCC) at the University of Edinburgh from September 2015 to August 2016. The Master's programme allowed us access to world class High Performance Computing platforms and technologies such as ARCHER, the UK’s fastest supercomputer. Other facilities included multicore, Xeon Phi and GPU machines.

“EPCC's MSc in High Performance Computing has always been a leader in its field. Coupling it to Data Science responds to the huge increase in demand for graduates with both HPC and Data skills from both science and business.”

Professor Mark Parsons
Executive Director, EPCC

The programme provided me with a thorough grounding in HPC along with a practical understanding of the fundamentals of data science. While the primary emphasis of the programme was High Performance Computing, it also included a significant Data Science element.

My dissertation, titled Performance Optimisation of LHC Workload at HPC Facilities was an application of HPC and Data Science techniques to Particle Physics. It focused on exploring the architecture of HPC systems in contrast to traditional clusters used in Grid computing, and explored opportunities for performance improvement of ATLAS Simulation jobs (at the LHC) when they were run on ARCHER, UK’s National Supercomputing Facility.

• It involved extracting and analysing unstructured data within log files generated by LHC (Large Hadron Collider) simulations to determine the cause of bottlenecks in the code when it was run at a large scale on ARCHER.
• While working towards optimising the working of the application to overcome these bottlenecks, I conducted an extensive study of existing literature on ATLAS simulation software and its implementation to HPC, and performed an elaborate testing process. 

The project was done as a collaboration between EPCC (Edinburgh Parallel Computing Centre) and the PPE (Particle Physics Experiment) group at the University of Edinburgh.