Sustainability in HPC

Abstract: Humanity is currently consuming energy at an unsustainable rate. As an industry HPC needs to help tackle this. From a research standpoint we have been part of the solution since the very beginning with climate modelling and energy research. However, the same effort has not been placed into making our installations as energy efficient as possible. In this talk OCF and Lenovo discuss ways in which HPC installations can be made more sustainable, both from a design and a technology perspective. It will include an initial presentation (supported by slides) which will cover hardware selection, scheduler configuration and cooling considerations as well as a customer case study.

​Introduction to the National Quantum Computing Centre (NQCC)​

Abstract: The National Quantum Computing Centre (NQCC) launched in September 2020 based upon an investment of £93m by UKRI over 5 years through EPSRC and STFC, as part of the National Quantum Technologies Programme (NQTP).   The aim of our work is to support UK government, industry, and the research community to enable the delivery of quantum computing capabilities for the UK.   Our focus is on the challenge of scaling Quantum Computers and to support the wider UK ecosystem.   This presentation will provide the background, the NQCC strategic intent, details about the NQCC facility (to be based in Harwell, Oxford), the technical programme, and a summary of the 5-year plan

EOSC and the EGI Cloud Federation​​

Abstra​​​ct: ​​The European Open Science Cloud (EOSC) is a European Commission initiative aiming at developing an infrastructure providing its users with services promoting open science practices. The EGI Fedeation is the largest distributed computing infrastructure in the world, and brings together hundreds of data centres worldwide and also includes the largest community cloud federation in Europe with tends of cloud providers across most of the European countries offering IaaS cloud and storage services. EGI has been actively contributing to the implementation of EOSC since its launch. This presentation will introduce the EOSC initiative and the EGI contributions to EOSC with a focus on the implementation of the EOSC compute platform and the delivery of various components of the EOSC Core services​

Expanding the role of HPC centres on training and collaboration for reproducibility

Abstract: At SC20 last month, I was invited to speak about my work and insights on transparency and reproducibility in the context of HPC. The title of my talk was "Trustworthy computational evidence through transparency and reproducibility," and the session's theme was Responsible Application of HPC. This talk will complement my SC20 invited talk by focusing in particular on the role of HPC centres. The question I wish to reflect on is: how could teams at supercomputing facilities work with researchers to help them adopt better reproducibility practices? 

Video of SC20 invited talk: https://youtu.be/pJiS0ReZynw

CoVal: A web-resource for validating cryo-EM structures of
SARS-CoV2 and mapping genomic mutations

Abstract: Since early 2020, a huge amount of research has gone into understanding the molecular mechanism behind SARS-CoV2 viral infection. Cryo-electron microscopy (cryo-EM) and tomography have been among the key techniques used to determine the three-dimensional structures of macromolecular components of the virus and how these components drive viral activity in the host. I will talk about our recent work on computational methods and tools for validating the cryo-EM structures of molecular assemblies of SARS-CoV-2 virus. The virus is evolving and its response to therapy may change with time. To understand how the SARS-CoV-2 genome mutations may affect structures and underlying molecular mechanisms, we have been developing an automated pipeline for identification of SARS-CoV2 genome mutations and map them onto 3D structures from cryo-EM. This will be deployed as an open-access web-service at the STFC cloud.

A Data-Centric Approach to Performance Portability

Abstract: The ubiquity of accelerators in high-performance computing has driven programming complexity beyond the skill-set of the average domain scientist. To maintain performance portability in the future, it is imperative to decouple architecture-specific programming paradigms from the underlying scientific computations. We show Data-Centric (DaCe) Python, which translates specially crafted Python code into parametric dataflow graphs. We demonstrate DaCe on CPUs, GPUs, and FPGAs over various motifs --- from fundamental computational kernels to graph analytics. We show that DaCe delivers competitive performance, allowing domain scientists to develop applications naturally and port them to approach peak hardware performance without modifying the original scientific code.