High performance computing is highly strategic and consists of combining a large number of processors – from several thousand to several million – to build parallel computing architectures and decrease computing time.
This race to acquire the highest computing power has high stakes for companies and governments: modelling climate change, optimizing energy production, analysing subsurfaces for petroleum exploration, financial analysis, technological simulation, cracking cryptography codes, applications for the defence sector etc.
In a world where data has become massive and complex, Inria teams make key contributions to developing programming and algorithms which are indispensable to hardware and software architectures that make it possible to achieve a very high level of computing power, as well as specific modelling and simulation tools for each field.
HPC and BigData
In its scientific plan for 2018-2022, the Institute has reiterated the importance of integrating the data dimension (whether instrumental or computational) into the HPC approach, and has decided to tackle the very current challenges of convergence between HPC and Big Data (HPDA - High Performance Data Analysis) approaches.
In this dynamic, Inria has created the "HPC - Big Data" Challenge. A structure brings together several project teams specializing in learning, Big Data and HPC, as well as partners such as Argonne National Lab, LBT/CNRS, ATOS/Bull and ESI Group.
Challenges on 3 axes of research
- The contribution of HPDA in large high-performance simulations (on the spot aspects)
- The contribution of HPC on a large scale for Big Data related applications such as Deep Learning
- The issue of infrastructure management and the different software stacks for resource management in a single, converged HPC framework - Big Data
Inria's project-teams contribute to the European ecosystem and maintain numerous collaborations, notably through their participation in a dozen collaborative projects in the Horizon 2020 program.
BDVA : The Big Data Value Association (BDVA) coordinates the European Public Private Partnership (PPP) on Big Data.
EuroHPC : This program, which aims to co-invest €1 billion in High-Performance Computing in Europe.
PRACE : Inria's implication in the PRACE program (Partnership for Advanced Computing in Europe), whose mission is to enable high-impact scientific research and R&D development in all disciplines to improve European competitiveness for the benefit of society, is carried out via GENCI, of which it is a partner.
(In French for the two first articles)