We would like to cordially invite you to participate at Grid Computing seminar held on October 15 at Masaryk dormitory, Thákurova 1, Praha. This seminar will be devoted to the overview of Czech national grid environment MetaCentrum describing its current status and implementation of new services in particular. The primary aim of this seminar is to inform current as well as new potential users of high-performance/high-throughput computing and about current possibilities available for solving of numerous research problems and challenges at national as well as international level. The seminar is co-organized by MetaCentrum and our partners from Bull company and we hope to provide a platform for useful discussions leading to the best resarch achievements for end users.
|Date||15. 10. 2010|
|Location||Masaryk dormitory, Thákurova 1, Praha 6 - Dejvice|
|Attendance||The attendance at the course is free (no fees); offered services are available only for academic public.|
|Registration & Logistics||
The registration is closed.
|Language||The course will be in Czech language except the speech of the guest of Bull company.|
|Schedule||Scheduled beginning at 10.00 a.m., expected end approximately at 5.00 p.m.|
|Program||10.00-10.15 X. Morise (French attaché for scientific and university collaboration): CZ/FR Universities Science Collaboration
10.15-11.00 Jean-Pierre Panziera: Bull - Invited lecture
11.00-11.10 GPU activities at FI MUNI and their results
11.30-12.30 MetaCentrum - Current MetaCentrum state, news - I, II, III
13.30-14.30 MetaCentrum - Practical guides, principles and ways of utilization
15.00-15.30 MetaCentrum - Collaboration of user communities - I, II
15.30-16.30 MetaCentrum - Future of national e-Infrastructure (CESNET, IT4Innovations, CERIT-SC)
16.30-..... Refreshment and free discussion
|Invited lecture annotation of foreign guest||Peta-scale HPC Applications Challenges
Jean-Pierre Panziera, Director of Performance Engineering, Bull Extreme Computing group
New large scale HPC systems with peak performance of 100s TFlops or even Petaflop(s) are regularly deployed. Often however, the applications performance difficultly scales to that level. The various system architectures with either Shared Memory or Distributed memory have both software and hardware limitations. The use of accelerators (e.g. GPUs) seems very promising but it is adding yet another level of complexity. This presentation reviews the challenges faced when scaling HPC applications to Peta-scale levels. Finally, technology trends which might shape the future Exa-scale systems are discussed.