ELEMENT - Exascale Mesh Network

Lead Research Organisation: University of Edinburgh
Department Name: Edinburgh Parallel Computing Centre

Abstract

The Exascale Mesh Network - ELEMENT - addresses the high priority use case of meshing for the Exascale (i.e. ensuring that meshes are of sufficient quality to represent Exascale problems and can be partitioned efficiently to minimise load imbalance) as well as meshing at the Exascale (i.e. creating highly scalable solutions that will be able to exploit extreme levels of parallelism).

The meshes required for Exascale simulations, under which we will aim to model problems with extreme geometric complexity and levels of refinement, will necessarily be very large with 10^9 cells and above, and contain cells that may differ in size by many orders of magnitude to faithfully resolve the underlying physics at their appropriate scales. Meshing and geometry management remain a significant bottleneck for complex applications on HPC platforms, posing a challenging obstacle that must be overcome to enable Exascale simulations. From a technical perspective, these issues include (but are not limited to) improved geometric handling, mesh adaptation and optimisation, intelligent meshing, automation and robustness, all within a large distributed environment that lies outside of our current capabilities.

ELEMENT's objectives are threefold: building a community around meshing practice by establishing a collaborative network; undertaking proof of concept studies, with prototype implementations of two target challenges; and publishing a Vision Paper and strategic research agenda covering the full meshing workflow, from mesh generation to adaptation, partitioning and visualisation.

Planned Impact

Many simulation applications from areas as diverse as engineering to astrophysics use meshes as a key part of their modelling technique. Overall, the economic and societal impact of such simulations is very high: every aeroplane we fly in, every car we drive in has been modelled in a variety of ways using a variety of meshes of different scales and complexity. ELEMENT focuses on understanding the broad set of applications that use meshes and how we will create and manipulate such meshes when using Exascale computers for our simulations. Improvements in the performance, scalability and efficiency of meshing will also reduce the cost of the simulations, both in terms of time and energy to solution.

The use of meshes spans a broad set of supercomputing applications, solving the problems of: mesh creation, mesh adaptation, mesh visualisation and mesh partitioning in parallel for Exascale systems and problems will enable high quality, high impact research in multiple areas of strategic importance. It is mandatory that we improve our ability to handle the scale, complexity and quality of the meshes that applications will require. ELEMENT will create knowledge impact through advancing the state of the art of the application if we are to properly take advantage of Exascale capabilities.

Programming Exascale systems successfully will require highly optimised applications and a strong focus on input and output data quality, which in turn require a new generation of supercomputing expert, from software engineers all the way to domain scientists. There is a severe shortage of people in the UK today with the necessary skills to design and deliver Exascale software that will scale to millions of parallel threads. ELEMENT will impact on people and skills development by identifying gaps and training the next generation of supercomputing simulation experts.

Publications

10 25 50