Massively parallel supercomputers of the “Blue Gene family” by IBM are nowadays among the most powerful and efficient (i.e. “environmentally sustainable”) HPC solutions. Starting from the first family member, Blue Gene/L, the goal of reaching operating speeds in the PetaFlops range, with low power consumption, has been achieved by assembling a very large number of cores (from some thousands to hundreds of thousands) with relatively low clock frequency (in the range of 850-1600 MHz), interconnected via fast network(s) for message passing. The underlying idea behind the Blue Gene design is therefore to support applications able to scale up to a very big number of tasks and/or threads, therefore allowing simulation of physical phenomena with continuously increasing size, resolution and/or degrees of freedom.

 

The present course intends to enable/support the scientific community to efficiently exploit the Blue Gene/Q system. More precisely, starting from a general overview concerning (massively parallel) supercomputers and HPC trends, the course aims at providing a full description of the Fermi configuration at Cineca, with special emphasis on main crucial aspects for users and application developers. For instance, details about compilation, debugging and optimization procedures will be provided, together with an overview of the available libraries, tools and applications currently available on the system. Examples of submission jobs will be discussed, together with scheduler (LoadLeveler) commands and queue definitions. A “practice session” is also planned.

• 23 January: Milan in Italian - SLIDES 
• 5 February: Rome in Italian
• 18 March: Bologna  in English. PATC course

Click on the date to register

Blue Gene architecture • Overview of Blue Gene/Q architecture and software (hardware components, network and partitioning, type of nodes and software stack) • Developing applications for BGQ (porting, debugging, spotting possible bottlenecks) • Running and monitoring jobs on BGQ (the runjob command, environment variables, job script examples, LL commands, Available Debugging and Profiling Tools)

1 dd.

For further information about HPC courses, contact corsicineca.it.

 

Cineca - BOLOGNA, Via Magnanelli 6/3, 40033 Casalecchio di Reno, Bologna ph. + 39 051 6171411
Cineca - MILANO, Via R. Sanzio 4, 20090 Segrate, Milan ph. +39 02 269951
Cineca - ROMA, Via dei Tizii, 6 - 00185 Rome ph. +39 06 444861