| Abstract |
Computational demands are growing in virtually every area of scientific
research and advanced engineering, from space, climate, and genetics
research, to automobile crash simulations and materials modeling, just to
name a few. The actual performance of many application programs, however,
doesn't seem to take a full advantage of the rapid evolution of hardware
and software technologies. The gap has been widening between the peak
performance of the LINPACK benchmark tests on the top-tier supercomputers
and the running performance of typical applications on average clusters.
Most applications were developed to optimally operate on old systems and
the migration to newer and faster systems seems taking tremendous time
and effort for most scientists. Still, with a proper support from
experienced experts, many existing scientific codes can be transformed
into scalable high-performance codes, effectively enhancing the
utilization of new HPC systems. As we foresee a peta-computing era in 5
years, therefore, the advanced user support becomes more and more
critical to sustain the growth of the effective computational power. In
this talk, I'd like to share ongoing challenges and issues that we are
facing now, from the real user experiences at Pittsburgh Supercomputing
Center.
|
