HighPerformance Computing Andrea Feltrin High Performance Computing and

  • Slides: 8
Download presentation
High-Performance Computing Andrea Feltrin High Performance Computing and Quantum Technologies DG CONNECT European Commission

High-Performance Computing Andrea Feltrin High Performance Computing and Quantum Technologies DG CONNECT European Commission

European HPC strategy in the c. PPP 2

European HPC strategy in the c. PPP 2

The new European HPC research landscape em tics M Exa. HYPE SAGE NEXTGen. IO

The new European HPC research landscape em tics M Exa. HYPE SAGE NEXTGen. IO NLAFET e ag or st Bio. Excel COEGSS Eo. Co. E Biomolecular Global systems Energy Mat h & ema y or Exa. FLOW Ex. CAPE Com. Pat Exa. NEST s m READEX h t i r Inte ESCAPE rcon go s ALLScale l l o o A t nec ing m t m a INTERTWINE HPC Progr Ecosystem ECOSCALE ANTAREX EXDCI Dat Exa. No. De Compute a-in Eurolab-4 -HPC tens Mont-Blanc 3 ive real -tim. Mango EXTRA e Green. FLASH Centres of Excellence E-CAM ESi. WACE MAX NOMAD Po. P Simulation Modelling Materials Performance optimisation Weather Climate Project Start: Autumn 2015

HPC in science and industry Material science Design of novel materials with desired properties

HPC in science and industry Material science Design of novel materials with desired properties (mechanical, functional, …) Computer Aided Engineering Fluid dynamics and structural analysis 4

HPC in the public sector Earth system modelling International negotiations 5

HPC in the public sector Earth system modelling International negotiations 5

6

6

The next challenge: exascale exa peta Performance parallel processing today tera integrated circuits massive

The next challenge: exascale exa peta Performance parallel processing today tera integrated circuits massive parallelism giga mega vacuum tubes transistors kilo 1960 1970 1980 1990 2000 2010 2020 Europe, USA, Japan and China are main global players in this race

Challenges to reach exascale • System power • Memory bandwidth and capacity are not

Challenges to reach exascale • System power • Memory bandwidth and capacity are not keeping pace with the increase in flops • Cost of data movement • Programming model will be necessary to hide the level of concurrency from applications • The I/O system at all levels will be much harder to manage • Reliability and resiliency will be critical at the scale of billion-way concurrency 8