NEWS BRIEFS
Dallas, TEXAS — At SC2000, PBS/Veridian and NASA’s Information Power Grid did a demo that will lead to allowing users of computational and data grids to reserve valuable computer resources for a time they will have a critical need.
A virtual computing environment that integrates computer resources at different sites – collectively known as NASA’s Information Power Grid (IPG) – is employed to simulate a large-scale memory and CPU-intensive aerospace application. This illustrates how IPG technology can be used to obtain a rigorous wake flow solution around the hovering rotors of a helicopter, where flow features are dominated by complex vortex dynamics.
The advanced reservation capability was demonstrated in the Argonne booth. This demo used Globus GARA and Veridian’s professional version of PBS, PBS Pro as well as MPICH-G2. This represents a major step forward for computational and data grids. Work will continue to make provide this capability for all grids using Globus.
Those involved: Dr. Jennifer Schopf who organized this effort; Adain Roy of the University of Chicago; Conrad Adbrecht-Buehler of Northwestern University; Nick Karonis of Argonne, working with MPICH-G2.
The Globus Project has developed numerous collaborations with members of the scientific research and development community. Many of our collaborators are joining us this year in demonstrating tools and applications that utilize Grid services provided by the Globus Toolkit.
MPICH-G2 is optimized for running on grids. It combines easy secure job startup, excellent performance, data conversion, and multiprotocol communication. However, when communicating over wide-area networks, applications may encounter network congestion that severely impacts the performance of the application. This demo demonstrates Quality of Service (QoS) extensions to MPICH-G2. By simply setting an attribute for a communicator, the application program can request quaranteed bandwidth from the network. With this extension, scientific programmers can easily ensure that programs that operate in busy networks will have consistent, high-quality performance.
============================================================