text-based alphabetical index Skip to content
StaffDirectory Search
Home
AboutNAS
ResearchNTechnology
CurrentResearch
Applications
ProblemSolvingEnv
Networks
HighPerformanceComp
MassStorage
grid common services
LongTermResearch
PapersNReports
Software
SampleDatasets
UserServices
ContactUs
Feedback
Sitemap
Help

Grid Common Services

Assessing Legion as Grid Middleware
Legion is an object-oriented software system that simplifies and systemizes access to grids of distributed computing resources. Using large distributed computational fluid dynamics simulations as benchmarks, NAS researchers are comparing Legion’s performance in the Information Power Grid environment with that of the Globus metacomputing toolkit.

Distributed CFD Using Globus
Researchers at NAS are attempting to run parallel, high-fidelity, three-dimensional computational fluid dynamics simulations on the distributed Information Power Grid computing network with as few changes to the code as possible. The Globus metacomputing toolkit is used to execute the code, and the researchers have devised a deferred timestep method that allows the overlapping of computation and communication, potentially increasing the speed of the simulation. In 1998, airflow around the X-38 Crew Return Vehicle was simulated using the CFD code OVERFLOW-D2 running concurrently on two SGI Origin 2000 computers at the NAS Facility and Argonne National Laboratory. A similar, but larger-scale computation was carried out in 1999 using SGI Origin 2000 computers at three geographically separated sites.

Globus Documentation
NAS technical writers have produced the
Globus Quick Start Guide (32-page 152K PDF) (To read this files you will need the free Adobe Acrobat Reader, The Globus Quick Start Guide is a brief introduction to using Globus software tools as a foundation for running jobs on the Information Power Grid. Among other things, the guide teaches users how to obtain Globus security certificates and how to run simple Globus jobs. As of this writing, no other in-print documentation is available. Collaborating with NAS on the documentation project are personnel at the National Center for Supercomputing Applications and the National Laboratory for Applied Network Research.

Globus Usability Studies
Under Globus 1.1, IPG users access the Globus metacomputing toolkit through an arcane set of command-line scripts and options (in the form of the Resource Specification Language, or RSL) that must be entered manually. The goal of Globus usability studies at NAS is to develop a graphical user interface that simplifies and automates the creation and submission of Globus scripts and RSLs. The GUI will eventually be adapted to work with batch queuing systems such as PBS, to speed the Globus authentication procedure, and to make data staging and job control nearly transparent to the user. This work is being carried out in collaboration with the Globus project at Argonee National Laboratory.

Grid Co-scheduling
Illustration of Globus distributed computing tools To use supercomputing resources at different sites at the same time, researchers need a system for co-scheduling and advance reservations. Staff working on the grid scheduling task at NAS are attacking this problem by deploying a full prototype advance reservation system across the NASA Information Power Grid network, currently running at Ames, Glenn, and Langley Research Centers. NAS is also working with the the Grid Forum to define standards for grid scheduling and resource management. Advanced grid scheduling capabilities will allow hardware owners to better manage their resources and will allow Grid users to better specify their priorities (for example, "lowest cost" or "overnight turnaround time.") This work is being carried out in collaboration with Argonne National Laboratory and the Information Sciences Institute at the University of Southern California.

Portable Batch System
The Portable Batch System (PBS) is a batch queuing system developed at NAS to span all of the Facility’s vector and parallel computers. Today, PBS is one core component of the Information Power Grid’s "middleware" layer, the set of tools that handle communication between heterogeneous, geographically distributed computing systems. NAS researchers are adapting the system to allow advance reservations of remote resources and to provide an easy-to-use interface to the Globus suite of distributed computing tools. In one of the NAS Systems Division’s best recent example of technology transfer, PBS is now being marketed commercially by Veridian Group, Mountain View, CA.

Curator: Jill Dunbar
Last Update: November 26, 2002
NASA Official: Walt Brooks