Personal-Computer-Based Data Acquisition System Developed
NASA Lewis has about 60 medium- to large-scale facilities for testing space and aeronautical structures, systems, and components--from full-scale jet engines to chemical rockets, multistage compressors, single-cell batteries, and simple stationary airfoils. The Scientific Data Systems group at Lewis has supported these test facilities with a scaleable set of minicomputer-based data systems for many years, using an evolved suite of on-line data-processing services. Also supporting these test facilities are numerous test stands and laboratories with less-demanding data acquisition requirements, and typically less funding for data acquisition systems. Over the past year we have developed a prototype, low-cost, personal-computer-based data acquisition system for use in small-scale test facilities.
For example, the NASA Solar Electric Propulsion Technology Application Readiness (NSTAR) Program's ion thruster is a state-of-the-art design thruster for satellite stationkeeping applications. It uses the ionization of a rare gas, such as xenon, to produce a very efficient low-thrust beam. An experiment was undertaken to demonstrate 2000 hr of continuous operation of the NSTAR thruster in a space vacuum tank at Lewis. The data acquisition requirements were to acquire, monitor, and record thruster voltages, currents, and gas flows and to provide open-loop control of a pair of thruster preheaters. The data acquisition hardware selected for this task consisted of a 16-channel analog-to-digital converter, a 16-channel analog multiplexer with channel-selectable signal conditioning, and two 4-channel digital-to-analog converters. All these components adhere to the IEEE 488 communications standards. This hardware was interfaced to a 486 DX2 personal computer through a general-purpose interface bus card. The software package LabVIEW for Windows, from National Instruments, was chosen as the programming language. LabVIEW, a graphical programming language, provides a graphical user interface with a data flow programming methodology.
Using LabVIEW, the following suite of services were programmed on the personal computer (PC):
For the NSTAR thruster life test the PC is performing these services at a steady 6-sec update rate. Data for the permanent record of the life test are being recorded once per hour. The history file contains a data record for every 2 min over the last hour and is frozen when an abort occurs. The PC can shut down the thruster in response to either low vacuum tank pressure or degraded performance of any of several thruster measurements. To date, the NSTAR ion thruster has logged over 1200 hr of operation. The total cost for the PC-based data acquisition system was less than $10,000, including the cost of the PC.
Lewis contact: Vincent J. Scullin, (216) 433-5156
Headquarters program office: OSAT
Asynchronous-Transfer-Mode Application Program Interface Developed To Support Parallel Virtual Machine
Two projects that use Fore Systems asynchronous-transfer-mode (ATM) interfaces required us to examine and possibly modify or develop the adapter and the application program interface (API) code. Both projects use ATM as a high-speed data communications medium in a distributed processing environment, such as the parallel virtual machine. In particular, we are examining the usefulness of ATM to support cluster-based computing and also as a fabric to provide high-performance communications between supercomputers and high-end workstation clusters. We need to determine if ATM will provide sufficient performance for the level of integration and interoperability required.
Both projects require not just high throughput but also low latency. Low latency is needed in a connectionless style of communications (as in personal or switched virtual circuits with rapid bandwidth reallocation capability) when the connectivity patterns are not known a priori, as well as in a more connection-oriented style. For example, in a parallel application written to execute on an Intel, the connectivity patterns are well known, and we could map a virtual topology to the ATM network so that the program could run without modification. Low-latency support is required in the context of a local environment. Geographically distributed computing will always have a high latency owing to the propagation delay, but we can often overlap this through problem partition.
Because we are using ATM for its high speed, we need to reduce the overhead that diminishes its performance. Much of the difficulty in terms of latency results from context switches and invoking the operating system through system calls. It is our understanding that the API was developed with the primary goal of increasing the maximum throughput. We intend to examine the API to see if the latency can be reduced, perhaps with a slight tradeoff in throughput, while also retaining the multiuser nature of the design.
We also want to examine a possible high-performance, single-user mode, which might be useful to a cluster, such as the Lewis Advanced Computing Environment (LACE), where general communication would be supported through the usual network (Ethernet, Fiber-Distributed Data Interconnect (FDDI), and Fiber Channel) and interprocessor communication would be supported over the ATM network. This configuration would significantly reduce API overhead but would constrain use of the interface to a single user. This secondary objective will be useful in specialized situations, such as with the LACE cluster, but will not be of general interest due to the strong constraint.
The first project using Fore System's ATM is an initial study of porting a version of a message-passing library, such as the parallel virtual machine or the message-passing interface, to communicate directly over ATM by way of the API. This capability will be needed because we expect that, once we have the API code and understand how it works, we will need to optimize use of the ATM API by the message-passing library to most effectively use this resource. Also, we will examine the possibility of having direct communication "hooks" in the adapter code. This particular application is usually characterized by initial broadcast traffic and then bursty and intermittent traffic.
The second project is a study of ATM as a communications medium between a supercomputer and high-end workstation clusters. The supercomputer generates numerical data, and the workstations are used for distributed computation, data fusion, and graphic visualization. Hardware inadequacies make it necessary to stripe the data over multiple network interfaces. To reduce the costs associated with this operation, we will examine the possibility of moving this striping function into the lower level code of the adapter or API.
Lewis contact: Jose M. Davis, (216) 433-5407
Headquarters program office: OA
Satellite and Terrestrial Network Applied to Engine Inlet Simulation
One of the ways NASA Lewis supports the High-Performance Computing and
Communications (HPCC) Program is by developing new technology that will help U.S. industry maintain its world leadership and then transferring this
technology in a timely manner. A technology that will help make the U.S.
aerospace industry more technologically and economically competitive is the
Numerical Propulsion System Simulation (NPSS), a "numerical test cell" that
will use high-speed computing and communications in place of conventional
testing.
Because the current communications networks connecting Lewis to its industrial collaborators cannot provide the necessary bandwidth for remote flow visualization, the HPCC Office is proposing use of the Advanced Communications Technology Satellite (ACTS) to link Lewis with remote industry sites. This concept will be tested by using ACTS to link Boeing and Lewis as they conduct a remote turbine engine inlet simulation in support of the High-Speed Civil Transport Program.
The purpose of this study is to prototype a "numerical wind tunnel" using high-speed computing and communications in place of a conventional wind tunnel. A series of numerical experiments will be conducted to develop a mixed-compression inlet control system, but these studies should be prototypical of any remote numerical analysis that requires near "real time" results. The inlet simulator will be executed on the Lewis Cray YMP supercomputer but controlled by Boeing in Seattle, Washington. Flow visualization information will be transported by ACTS to the remote site for quasi-real-time display on high-performance, three-dimensional graphics workstations.
The experimental ACTS satellite network will allow aerospace manufacturers to remotely access NASA's distributed engine inlet simulation application. Synchronous Optical Network/asynchronous transfer mode (SONET/ATM) traffic studies will be performed on the experimental network to evaluate its performance in the NASA engine inlet simulation application. The network's performance for this application will be optimized experimentally and then compared with the performance of an all-terrestrial, long-haul SONET/ATM network for this application.
Bellcore's High-Speed Switching and Storage Technology Division will participate with NASA Lewis in systems engineering and traffic studies on the experimental high-speed SONET/ATM test bed. This test bed operating at both the 622- and 155-Mbps rates will interconnect a Cray YMP supercomputer at NASA Lewis to high-end workstations at Boeing.
This experimental network will be among the first integrated demonstrations of numerous advanced high-speed technologies in support of state-of-the-art aeronautical and network research.
Lewis contact: Isaac Lopez, (216) 433-5893
Headquarters program office: OA
Heterogeneous, Geographically Dispersed ATM-Based Distributed Computing Assessed
NASA Lewis has experimentally assessed the coupling of two emerging fields:
asynchronous transfer mode (ATM) and cluster-based computing. Cluster-based computing has become an area of serious interest owing to the large numbers of commonly available high-end workstation processors. The goal is to harness the enormous number of idle computing cycles for use in a distributed application. The economic factors and the risk amortization, when compared with the purchase of a supercomputer system, make this approach extremely attractive.
However, cluster-based computing requires a network to support low-latency interprocessor communication. Although approaches such as Fiber-Distributed Data Interconnect (FDDI) have been developed for this environment, ATM is particularly attractive mainly for its low-cost, scaleable performance and its potential support of geographically distributed computing. Low cost is essential. One reason why ATM has attracted so much interest in this environment is its initial lower cost compared with FDDI interface cards when they were first introduced.
This work uses an ATM test bed to evaluate the performance implications of flow creation and warm-start behavior. It also identifies the latency and throughput characteristics achievable via the traditional protocol stack of transmission control protocol/Internet protocol (TCP/IP), with both ATM adaptation layer (AAL) 3/4 and AAL 5. The overhead of the message-passing library, such as parallel virtual machine (PVM), which uses the user datagram protocol (UDP), is examined. The NASA numerical aerodynamic simulation (NAS) benchmark suite is used to study application-level performance implications of the processes communicating through PVM on the ATM test bed. The operating system interface to PVM is then modified to circumvent UDP/IP and allow direct communication between the adaptation layer and the message-passing library to assess the performance implications of eliminating the transport layer protocol overhead together with the additional buffer copies.
Lewis contact: Russell W. Claus, (216) 433-5869
Headquarters program office: OA
High-Performance Computing and Communications K-12 Project Supports Schools
NASA Lewis began the K-12 project to inspire students from kindergarten through high school to pursue careers in science and engineering. A particular focus is on underrepresented schools and minorities.
We have involved teachers early on and continually throughout the project, making training available to raise all participants to the same level of expertise and to standardize on computing platforms so that all participants could easily share advances within the project. The K-12 project focuses on three areas of development:
We have trained 25 teachers from 14 schools, high school and elementary, and have given nine schools Apple MacIntoshes and network equipment for connecting to Internet. The two-week teacher training conducted each summer consists of instruction by Lewis personnel on topics including Mac Basics, Internet, visualization, computer languages, Unix, Interactive Physics, Maple, Animation Works, and Spyglass. One result is that Barberton High School will teach a new course entitled High Performance Computing at the 10th and 11th grade level. We have implemented customary and innovative network efforts within the K-12 project. Support for connections to Internet ranges from basic telephone line access to a successful implementation of radiofrequency technology at sustained T1 speeds (1.5 Mbps). Cleveland East Technical High School has partnered with Cleveland State University to acquire Internet access and to demonstrate the cost-effective use of this "wireless" communications path.
The K-12 project inspires students, teachers, and NASA personnel to develop and enhance school curricula into living entities that can grow and accommodate the technology already available outside the classroom. The program will continue to provide two weeks of teacher training annually, computers to selected schools, and basic Internet connections. In fiscal 1995 we propose working with the sight impaired and developing the Lewis Teacher Resource Center into a functioning instructional facility for year-round K-12 use.
Lewis contacts: Gregory J. Follen, (216) 433-5193; Gynelle C. Mackson, (216) 433-8258
Last updated 1995
Responsible NASA Official:
Gynelle.C.Steele@nasa.gov
216-433-8258
Point of contact for NASA Glenn's Research & Technology reports:
Cynthia.L.Dreibelbis@nasa.gov
216-433-2912
SGT, Inc.
Web page curator:
Nancy.L.Obryan@nasa.gov
216-433-5793
Wyle Information Systems, LLC