ARCHER News

Tuesday 29th October 2019

Enabling multi-node MPI parallelisation of the LISFLOOD flood inundation model

Wednesday 6th November 2019 15:00

Arno Proeme, EPCC

The availability of higher-resolution topographic datasets covering greater spatial domains has soared in recent years, pushing the limits of computational resources beyond those typically found in regional HPC services. In addition, many countries that are subject to large-scale flooding each year do not have access to real-time flood forecasting software. This webinar describes how HAIL-CAESAR, a geosciences code that implements the LISFLOOD flood inundation model, was ported to make use of LibGeoDecomp - a C++ stencil code HPC library - to enable multi-node parallelism. Whilst currently single inundation scenarios can take multiple days to run using standard hydrological modelling software, this project paves the way for ensemble runs that can be initiated on the basis of a 24 or 48 hour rainfall forecast and complete within shorter timescales, which should ultimately have major implications for flood warnings in developing countries.

Full details and join link http://www.archer.ac.uk/training/index.php

A Fully Lagrangian Dynamical Core for the Met Office/NERC Cloud Model - Online webinar

Wednesday 13th November 2019 15:00

Steven Boeing, University of Leeds and Gordon Gibb, EPCC

We discuss the development of an essentially Lagrangian model of atmospheric cloud dynamics, carried out in the eCSE project ‘A fully Lagrangian Dynamical Core for the Met Office/NERC Cloud Model’. We describe the work, which aimed to incorporate the ‘Moist Parcel in Cell‘ (MPIC) code into the Met Office/NERC Cloud Model (MONC), thereby granting it MPI parallelism and the ability to scale beyond a single node.

In the Moist-Parcel-In-Cell (MPIC) model, parcels represent both the thermodynamic and the dynamical properties of the flow. The representation of parcel properties is fully Lagrangian, but an efficient grid-based solver calculates parcel advection velocities. The Lagrangian approach of MPIC has a number of advantages: in particular, parcel properties are naturally conserved, and the amount of mixing between parcels can be explicitly controlled. MPIC also lends itself well to parallelisation, because most of the communication required between processors is local, and an efficient solver is available where global communication is required.

First, we describe the modifications to MONC in order for it to be parcel-aware. Subsequently, we outline the implementation of MPIC’s physics into MONC. We then provide an in-depth comparison of MPIC with the newly developed code, and investigate how it scales to many thousands of MPI processes. Finally we discuss the limitations of the code, and future work to be carried out to mitigate these limitations. We found that the code can scale up to many thousands of cores for large problem sizes, although the main limiter of performance at scale are the Fourier Transform routines. Despite good MPI performance from the code, OpenMP performance is poor, achieving a speedup of only 3 on 12 threads. Overall, on a single node the new code performs better than MPIC does, carrying out more parcel operations per core per second.

We will also describe how users can install the code and develop new components.

Full details and join link http://www.archer.ac.uk/training/index.php

Free "Supercomputing" MOOC started on 14th October 2019

www.futurelearn.com/courses/supercomputing/.

Today's supercomputers are the most powerful calculating machines ever invented, capable of performing more than a thousand million million calculations every second. This gives scientists and engineers a powerful new tool to study the natural world - computer simulation.

This free 5-week online course will introduce you to what supercomputers are, how they are used and how we can exploit their full computational potential to make scientific breakthroughs. Register for the upcoming run on 14th October at www.futurelearn.com/courses/supercomputing/.

The FutureLearn platform has a paid "Upgrade" option that gives you access to the material for an extended period and the ability to obtain a Certificate of Achievement based on online tests. However, all other parts of the MOOC can be accessed for free during its run.

This course was developed by EPCC at the University of Edinburgh and by SURFsara as part of the EC-funded PRACE project.

HPC-Europa3 Transnational Access programme

Collaborative research visits using High Performance Computing

Call for applications: next closing date 20th November 2019

HPC-Europa3 funds research visits for computational scientists in any discipline which can use High Performance Computing (HPC).

Visits can be made to research institutes in Finland, Germany, Greece, Ireland, Italy, the Netherlands, Spain, Sweden or the UK.

UK-based researchers can benefit in two ways: either by visiting a research group elsewhere in Europe, or by hosting a research visitor from another country.

What does HPC-Europa3 provide?

  • Funding for travel, living and accommodation expenses for visits of up to 13 weeks.
  • Access to world-class High Performance Computing (HPC) facilities.
  • Technical support to help you make best use of the HPC systems.
  • Collaborative environment with an expert in your field of research.

Who can apply?

  • Researchers of all levels, from postgraduate to full professors.
  • Researchers from academia or industry.
  • Researchers currently working in a European Union country or Associated State (see http://bit.ly/2PkVsSV for full list of Associated States).
  • Researchers may not visit a group in the country where they currently work.
  • A small number of places are available for researchers working outside these countries - please contact staff@hpc-europa.org for more information.

How do I apply?

Apply online at http://www.hpc-europa.org

The next closing date is 20th November 2019. Closing dates are held 4 times per year. Applications can be submitted at any time. You should receive a decision approximately 6 weeks after the closing date.

For more information and to apply online, visit: http://www.hpc-europa.org/

Follow us on Twitter for project news: https://twitter.com/HPCEuropa3

Upcoming Training Opportunities

Registration open now

  • Hands-on Introduction to HPC for Life Scientists, Birmingham, 30 Oct - 1 Nov 2019
  • Free "Supercomputing" MOOC starting on 14th October 2019
  • Enabling multi-node MPI parallelisation of the LISFLOOD flood inundation model, Online webinar, 6 November 2019 15:00
  • A Fully Lagrangian Dynamical Core for the Met Office/NERC Cloud Model, Online webinar, 13th November 2019 15:00
  • Shared-memory Programming with OpenMP - Online course, four Wednesday afternoons, 13, 20, 27 November and 4 December 2019

Registration opening soon

  • HPC Carpentry, EPCC, Edinburgh, 9 - 10 December 2019

Full details and registration at http://www.archer.ac.uk/training/index.php