12/28/01 -- NAS Scientist Co-develops Unique Load-Balancing Solution
NAS Division scientist Rupak Biswas has codeveloped a novel load-balancing algorithm for scientific applications. The method is described in "Parallel Processing of Adaptive Meshes with Load Balancing," which appears in the December 2001 issue of IEEE Transactions on Parallel and Distributed Systems (TPDS). Biswas collaborated with Sajal Das and Daniel Harvey, University of Texas at Arlington.

The development team's goal was to create a load-balancing algorithm that also minimized interprocessor communication and data redistribution delays for unstructured grids.

Many scientific applications involving grids lack a uniform underlying structure. These applications are often also dynamic in nature in that the grid structure changes significantly between successive phases of execution on computers.

Mesh adaptation of unstructured grids through selective refinement and coarsening has proven effective in solving complex computational problems. However, an approach using parallel processors is complicated because of frequent load-balancing requirements. Traditional dynamic load balancers are for the most part inadequate because they lack a global view of system loads across processors. The team's paper describes a general-purpose load balancer that utilizes symmetric broadcast networks (SBN) as the underlying communication topology.

Comparisons with PLUM, a successful global load-balancing framework, shows that this SBN-based load balancer achieves lower redistribution overhead for dynamic irregular applications by overlapping processing and data migration.

For more information on the SBN-based load balancer, contact Rupak Biswas at rbiswas@nas.nasa.gov.

12/20/01 -- IPG Workshop Highlights Collaboration, Web Services
Researchers from across the U.S. gathered recently to discuss the current state of NASA’s geographically distributed computational network, the Information Power Grid (IPG). At the third annual IPG workshop, held December 4-5 in Palo Alto, Calif., more than 100 computer scientists from academic institutions, government labs, and NASA centers learned about recent developments in grid technology.

Thirty scientists presented the state of their research and the development progress of IPG infrastructure components. Peer feedback, problem solving, and networking allowed researchers to address problems encountered in developing new Web tools and applications for the IPG.

Kicking off the two-day event, NASA Advanced Supercomputing (NAS) Division Chief Bill Feiereisen emphasized the importance of collaboration. “I think it is important to realize that even though this is NASA’s IPG workshop, the IPG is part of an entire grid effort throughout the United States, and throughout Europe, and the Asian Pacific Rim — I want to make it really evident that we are doing this all together. Many levels of many different organizations are working toward the same goal of merging efforts into a single grid.”

Although presenters came from various institutions and reporting on different topics, there was a central theme and direction to all presentations. “It’s pretty clear that the area of Web services is going to be the hot topic this year in grids. There was quite a bit of discussion in all of the talks about what Web services are,” said Tony Lisotta, NAS Division IPG task leader.

Among the presenters: Geoffrey Fox, Indiana University, who discussed the need to define interfaces for web applications to facilitate computer-to-computer interactions in “Peer to Peer Networks and Web Services for a Community Grid”: George Myers, NASA Ames Research Center, outlined recent developments in LaunchPad, a Web portal created to enable grid users to submit jobs to IPG resources via any Web browser; and Robert Griffin, NASA Glenn Research Center, discussed interactive online job submission for monitoring and modeling aircraft through the Aviation Safety Program.

One outcome of the workshop: new collaboration was formed at the workshop between NASA’s IPG and the National Virtual Observatory (NVO) project. The NVO project aims to make available to individual astronomers images of the night sky captured with powerful telescopes. NVO would like to provide high-powered computational resources necessary to process the data. It is this sort of collaboration that enables the IPG to grow and expand its capabilities.

NAS Division senior scientist Tom Hinke summed up the event: “I thought the workshop was extremely successful. We had good participation from the general grid community this year, and I hope to see more applications people at future workshops.”

For more information on NASA’s Information Power Grid workshops, contact Tom Hinke at thinke@mail.arc.nasa.gov.

12/07/01 -- Cart3D Used to Study Man-Portable Surface to Air Missiles
Scientist in the NAS Division's Applications Branch have extended the capabilities of NASA's Cart3D aerodynamic analysis software to simulate the unsteady flight of infrared-guided, man-portable surface-to-air missiles (SAMs). The simulations are being used to analyze the aerodynamic performance characteristics of these SAMS.

In the late 1970s, these portable SAMs proved to be highly effective weapons against the Soviets in Afghanistan. Since then, these weapons have been sold by the thousands on the black market to various terrorist organizations around the world. While they are very inexpensive to purchase, these missiles have unique flight characteristics that make them difficult to simulate computationally.

Michael Aftosmis and Scott Murman conducted steady and unsteady simulations of controlled flight for representative SAMs on both the 1,024-processor and 512-CPU SGI Origin 3000 systems at the NASA Advanced Supercomputing facility at Ames Research Center. Over 720 steady simulations were run on meshes totaling almost 2.4 billion cells.

In 2000, the Defense Intelligence Agency (DIA) Missile and Space Intelligence Center enlisted the help of NASA and other research organizations to numerically simulate such missiles. The information gained will enable the DIA to develop effective measures for countering the threat to civilian and military aircraft posed by terrorists armed with such missiles.

"Results from preliminary NAS studies presented to the DIA have been met with enthusiastic support," said Aftosmis. The steady and unsteady results have shed new light on the flight characteristics of these missiles. Analysts at DIA have relied on early versions of NASA's Cart3D package for several years, and they have already begun to incorporate results from the preliminary NAS Division studies with their in-house studies. Analyses like these allow DIA personnel to plan effective evasive maneuvers for aircraft targeted by these missiles, and support other counter-terrorist work.

The Cart3d package permits end-to-end simulation of inviscid three-dimensional flows around complex vehicles. The parallel analysis code in Cart3D ("flowCart") was developed under NASA's former High Performance Computing and Communications Program, and is well suited to these computations because it scales very well on parallel shared-memory hardware. In addition, many computations can be launched simultaneously to rapidly fill out test matrices. Cart3D was developed at Ames and New York University and is available through the NASA Ames Commercial Technology Office.

For more information on Cart3D, contact Michael Aftosmis at aftosmis@nas.nasa.gov.

10/30/01 -- NASA IPG Team Meets Another Major Milestone
NASA's Information Power Grid (IPG) team reached another major project milestone by demonstrating the use of grid services for remotely connecting to scientific instruments, as well as distributed, real-time access to instrument data.

The purpose of these demonstrations was two-fold: to show that the IPG services and infrastructure can provide scientists and engineers with on-demand connectivity to high-data-rate instruments and an advanced collaboration environment; and to show that IPG resources can be used to store and access data collected from one or more remote instruments.

The IPG is designed to take a large collection of dispersed and heterogeneous resources -- computing systems, storage systems, and scientific instruments -- and define a standard set of services for accessing those resources for research. NASA's Ames, Glenn, and Langley Research Centers, as well as its Jet Propulsion Laboratory are collaborating to develop infrastructure for the grid.

The team demonstrated two scenarios. First, using the wind tunnel instrumentation system (called DARWIN/DREAM), the team showed that NASA instrumentation facilities can use IPG services and resources in order to: give the instrumentation facility access to large-scale computing and data systems; provide a standardized set of highly capable access and management services that do not have to be developed by the application developers; and provide wider access to instrumentation data and analysis systems through a widely deployed IPG.

In the second scenario, using the UC San Diego, TeleScience for Advanced Tomography facilities, the team demonstrated a “use-model” where NASA researchers using remote, non-NASA facilities, could use NASA IPG services and other resources to both store and analyze data obtained at remote instrumentation facilities.

All of the critical data paths for both demonstrations transferred data at 50 megabits per second or greater.

For more information, contact Arsi Vaziri at vaziri@nas.nasa.gov.

10/4/01 -- NAS Speeds New Rocket Turbine Design
Engineers at NASA's Marshall Space Flight Center (MSFC) were recently able to shave three-and-a-half months off the task of simulating a supersonic turbine, thanks to the work of NAS Division user services staff. The simulations were performed on supercomputers located at the NASA Advanced Supercomputing (NAS) Facility, to analyze the unsteady flow in an advanced turbine design. The results will help in the design of more efficient and durable rocket engines.

Responding quickly to a request from MSFC's Applied Fluid Dynamics Analysis Group, Chuck Niggley, NAS Scientific Consulting Group lead, and colleague Herbert Yeung, provided access to special computational resources when regular resources proved insufficient for the MSFC simulation task. The NAS team set up a special account for users, and worked around the clock to keep MSFC apprised of anticipated changes in the computing environment.

In a letter of thanks to NASA officials, Daniel J. Dorney, of MSFC's Applied Fluid Dynamics Analysis Group, said: "Both Chuck and Herbert should be complimented on their willingness to help users solve computer-related issues, their understanding of users deadlines, and their positive attitude."

MSFC has designed and experimentally tested a supersonic turbine, called the "Simplex" turbine. This "partial-admission" turbine (that is, the flow enters the turbine over only a portion of the turbine inlet ) was designed as a ground demonstrator, and was experimentally tested with both metal and composite airfoils (blades) to study the feasibility and durability of advanced composite materials. "The experiments showed that the composite airfoils performed very well," said Dorney.

While the experiments yielded significant insight into the flow physics, an assessment of the unsteady flow field required the extensive use of three-dimensional simulations. The simulations were also intended to provide a better understanding of the flow physics of
partial-admission turbines, often used in rocket engines . According to Dorney, the simulations were also used validate NASA's CORSAIR code for partial admission and
full-annulus simulations.

Although computational models normally consider only a portion of the 360-degree annulus of the turbine, the nature of the Simplex turbine required simulation of the full 360 degree annulus, modeling all of the airfoils, Dorney explained. The computational grid used in the simulations contained in excess of 7 million grid points and required approximately two months of wall clock time for each simulation.

Specific goals for these simulations were to determine the unsteady pressures experienced by the turbine airfoils to help interpret the experimental data and to help determine the durability of the airfoils. Another goal was to determine the effects of a partial admission flow field on the performance of the turbine. The simulations served as precursors for many Space Launch Initiative tasks, which consider the use of partial admission turbines.

For more information about computational resources at the NAS Facility, contact Chuck Niggley at niggley@nas.nasa.gov.