Cybershake on opportunistic cyberinfrastructures

Continued improvements in provisioning capabilities of cyberinfrastructures
increase the computational enablement of science such as the Cybershake
application. Using workflow tools such as Swift and Pegasus, we observed how
jobs are scheduled to exploit an expanded set of resources available in TeraGrid
and Open Science Grid (OSG). Our tests on OSG gave 2000~CPUs in a day on an
opportunistic basis. This was well within the time-to-solution requirements of
the Cybershake application. However, there were potential bottlenecks in the
data requirements of the application. In this paper, we characterize how the
time-to-solution budget is spent by looking at the running time and file
transfer time of the post-processing computation.