This proposal is the result of a recent visit by key faculty and university administrators from Virginia Tech with NCSA staff, following an invitation from NCSA to be part of the NCSA-SGI Power Grid Alliance Program (http://www.sv.vt.edu/future/NCSA-VT_visit.html). Recent programs in supercomputing and visualization at both campuses have evolved to a point where collaborations on topics of shared interest could benefit both organizations' short and long term goals. This proposal is consistent with Virginia Tech's commitment to enhancing its teaching and research mission through the use of supercomputing, human-computer interfaces and visualization in various research programs, undergraduate and graduate curriculum, and through extension to Virginia as a land grant university.

Access and use of the existing SGI-Power Challenge in the College of Engineering will be documented along with voluntary performance bench marks that will be compared with the other existing parallel computing resources on campus. Our short term goals will be to provide bench mark results that will be voluntary and compiled by a collaborative web page. This web page will be designed to query and catalogue daily activities by the use of web forms. At logon and at the end of each batch job the users will be queried and asked to participate. We will especially encourage users who use the same code on two different parallel machines to participate in the program. Long term goals are: 1. study the use of MPI and PVM in linking resources via a high bandwidth network and 2. to track these activities as part of our planned Advanced Communications and Information Technology Center (ACITC) building (http://www.sv.vt .edu/future/future.html#ACITC).

The HCI Center, Laboratory for Scientific Visual Analysis and key faculty in the Colleges of Architecture, Arts and Sciences, and Engineering are particularly interested in the creation of a CAVE resource that will enhance existing educational and research programs on campus. Virginia Tech proposes to cost share in the purchase of a SGI Power Onyx with at least two VR boards by submitting a Academic Research Infrastructure (ARI) proposal to NSF; a white paper is due December 15, 1995. Our long term goals are to extend this type of resource as part of the ACITC. After several discussions with faculty throughout the University we discovered that CAVE technology has broad based support and will be used as a multidisciplinary tool. Professors across campus are as interested in using CAVE technology for the classroom as they are for research. Hence this proposal fits nicely with the University teaching and research mission. We are particularly fortunate to have a HCI group which will be a significant team component. Depending on the level of outside funding and matching University funds, the HCI group will facilitate the creation of general three dimensional (3D) interfaces that will benefit educational and research programs equally.

Our goal is to create a visual computing environment by linking the proposed CAVE and existing parallel computers on campus. To facilitate the link between existing numerical computing and the proposed visual computing environments, the University began the process of upgrading to an ATM campus network in spring 1995. This network can provide 155-Mbps links (OC-3) between parallel computers as well as individual workstations. We will also study the use of MPI and PVM to link parallel computers with the CAVE on campus. Future plans call for a state-wide ATM network that can encourage alliance interaction with industry, government, and other academic institutions in Virginia.

A very important component to all of these efforts will be a faculty-staff exchange program where individuals will benefit by a short term visit to NCSA from Virginia Tech and vice versa. We suggest that such visits be arranged by individuals at the respective institutions and motivated by shared interest on projects where there is obvious overlap and collaborations that can benefit both institutions' programs (e.g., development of CAVE technology). Our educational programs are also growing where we already have several courses on high performance computing, computer graphics, and network computing that visitors may want to take during their visit. There has been some discussion on organizing these courses into a minor on a topic that would be of interest to our allianace and industrial partners. It is agreed that visitors will pay for their own expenses but the host institution will provide office space, supplies and access to computing resources.

Recently the University has acknowledged the growing interest in supporting on- and off-campus activities in information technology by the creation of the Human-Computer Interaction Center. When combined with the existing Blacksburg Electronic Village (BEV), (http: //www.bev.net), the Human Computer Interface Resource Facilities, (http://pixel.cs.vt.edu/~rreaux /hci/research/facility.html), the Laboratory for Scientific Visual Analysis, (http://www .sv.vt.edu), the Multimedia Laboratory (http://www.multimedia.vt.edu) and existing hardware resources (SGI-Power Challenge, Intel Paragon, and IBM SP2), Virginia Tech is in a unique position to be a model alliance partner.

Virginia Tech is also a charter member of the Virginia/ICASE/Langley Program in High Performance Computing and Communication (VILap-HPCC; see http://www.cs.odu.edu/HPCC/hpcc.html). VILaP is a fellowship program that supports students working towards the Ph.D. degree in Computer Science or related disciplines, whose dissertation topics include a strong emphasis on applications of interest to NASA. The purpose of the program is to enhance NASA's ability to conduct computationally-intensive aspects of its research mission, and to educate applied computer scientists capable of lifelong contributions to "Grand Challenge" problems.

Existing NSF research programs that would participate in the alliance are:

1. the NSF Science and Technology Center on High Performance Polymers and Adhesives (http://www.vt.edu: 10021/research/stc/STC.Home.Page.html). In particular, notice that one of the four Research Thrust Areas (http://www.vt.edu:10021/research/stc/STC.Research.html), Resin Transfer Molding, has, in the past, required extensive supercomputing as well as visualization resources. Dr. Al C. Loos, who is responsible for this thrust area, has expressed an interest in working with the alliance: 1. to access supercomputing resources at NCSA and 2. continue to work on developing visual data analysis techniques to interpret model predictions. In return he will provide benchmark summaries of model predictions. Drs. Loos and Kriz recently participated in a NCSA-NIST sponsored workshop on "Modeling the Development of Residual Stresses During Thermoset Composite Curing". Future alliance collaborations will include organizing similar workshops on topics appropriate to the alliance. Industrial partners, associated with the NSF-STC, will be encouraged to participate in alliance programs.

2. a new research project in the Computer Science Department, sponsored by the NSF Network Infrastructure for Education program, "Leveraging Networks for Collaborative Education in the Blacksburg Electronic Village (BEV)." This project will build on the existing network infrastructure established by the BEV, using participatory design methods to develop and evaluate virtual lab materials used collaboratively by students in several area middle and high schools. We hope that the alliance will facilitate access to the network-based educational and collaborative technology under development at NCSA, and provide critical technical support in creating rich interactive network-based user interfaces. In turn, we will provide teacher- and student-centered design requirements for the continuing development of network-based collaboration facilities, and an in-depth evaluation of this technology in use in a real world setting.

3. another NSF program, for Education Infrastructure, improving learning in computer science through interactive technologies and digital libraries. In connection with this, a joint venture of the University and IBM will involve a 4 processor IBM SMP facility with 2 terabytes of hierachically managed storage scheduled to arrive late 1996, which will run IBM and locally developed digital library software. In addition to supporting computer science information from associations like ACM, this facility will afford access to other collections of text and multimedia information, provided by commercial, association and local (e.g., faculty) publishers.

1. Scientific Visual Data Analysis and Multimedia, ESM4984, by Professor Ron Kriz and Mr. Gordon Miller, is an undergraduate course with graduate credit that explores the use of supercomputing models and visual data analysis of the results in a variety of topics in engineering and the sciences. A summary of the class projects is posted on the Web at http://www.sv.vt.edu/class/ESM5984.html.

2. High Performance Scientific Computing, UH3004 (Undergraduate Honors Course), by Professors Cal Ribbens and Chris Beattie, this course would benefit by access to parallel computing off campus. A Web page can be found at http://www.sv.vt.edu/class/UH3004.html.

3. Molecular Modeling of Proteins and Nucleic Acid, BION5984 (Web format in progress), by Professor David Bevan, is a graduate-level course in which students are introduced to the theory and practice of molecular modeling techniques. Topics include molecular mechanics, molecular dynamics, quantum mechanics, free energy simulations, docking, drug design, and homology modeling. Emphasis is placed on practical aspects of modeling, with all students being required to perform and present a molecular modeling project. Computing and visualization tools associated with the alliance would be very important resources for this course.

4. Introduction to Human-Computer Interaction, CS3724, by Professor Mary Beth Rosson, introduces advanced undergraduates to a range of user interaction techniques, their rationale, and methods for designing and evaluating user interfaces. A key aspect of the class is demos of state-of-the-art user interfaces, and the alliance would enhance this considerably. A Web page can be found at http://ei.cs.vt.edu/~hciintro .

5. Computer-Supported Cooperative Work, CS5734, by Professor Mary Beth Rosson, is a new graduate course that will introduce techniques for analyzing and designming collaborative applications. NCSA experience and applications in this area, as well as the CAVE we hope to create at Virginia Tech, will be invaluable resources in this course

6. Computer Simulation Techniques in the Properties and Structures of Materials, MSE5984, by Professor Diana Farkas. In this course the student learns how to use model interatomic potential, molecular statics and dynamics, and Monte Carlo techniques to study surface phenomena, interface structure, diffusion, plastic deformation, and fracture.

7. Computer and Nework Architecture II (Networks) EE/CS 5516, by Professor Scott Midkiff, is a graduate-level course on computer networks. The presentation is in the context of the Open Systems Interconnect (OSI) seven-layer reference model, with particular focus on the data link, network, and transport layers. The course is intended to provide students with an understanding of fundamental concepts in networks and protocals and an ability to analyze networks and protocols that transcend any particular standard.

8. Telecommunication Networks, EE4984, by Professor Scott Midkiff, is an undergraduate course that is designed to give students an understanding of the architecture, technology, operation, and evolution of telecommunication networks. This includes networks for voice telephony, data and itegrated services.

9. Multimedia, Hypertext and Information Access, CS4604 (temporarily 4984) and Information Storage and Retrieval taught by Professor Edward Fox, both cover multimedia and hypermedia technology and can benefit from the CAVE and other facilities.

These courses, among others, can contribute to and benefit from the alliance. Our long term plan is to organize some of these courses into a minor in supercomputing, computer graphics, and computer networking.

• SGI Power Challenge (http://www.eng.vt.edu/eng/computer/crunch.html), 4 CPUs, 640MB memory, 18GB disk.

• Intel Paragon, XP/S-7, 100 General Purpose nodes where each node has two Inteli860 XP processors one dedicated to applications and the other to message passing, 32 MB memory, performance rated at 100MFLOPS single precision and 75MFLOPS double precision.

• IBM SP2 (http://spider01.cc.vt.edu/), 2 wide nodes with 512MB memory and up to 13GB disk and 8 thin nodes with 128MB memory and up to 4GB disk; this will be upgraded with at least 4 more nodes late in 1996.

Future efforts will target the construction of a CAVE environment with the purchase of an SGI Power Onyx with at least two VR boards. This effort will be cost shared by Virginia Tech together with the submission of a proposal to the NSF Academic Research Infrastructure (ARI) Program. The present SGI Power Challenge is targeted for numerical simulations only, hence, a visual computing environment such as a CAVE will complement the existing resources. We are particularly interested in network links with existing resources both on- and off-campus. In particular we are interested in learning how to link the CAVE visual computing environment with our SGI Power Challenge and other parallel computing resources on campus to create a "realtime" visual computing environment using an advanced communications network on campus.

• Intel Paragon,: The 28 node paragon was installed in June of 1994. Another 72 nodes were donated by NASA Langley, late 1996. It has been used by over 40 faculty and graduate students on campus for reserach in a wide range of areas, including multidisciplinary optimization, composites, numerical linear algebra, tomography, numerical methods for partial differential equations, and parallel discrete even simulation. The machine has also been used by several undergraduate and graduate courses in parallel computation, numerical analysis, and high performance computing.

• SGI Power Challenge: *

• IBM SP2: *

* both the SGI Power Challenge and the IBM SP2 are less than a year old and it is not possible at this time to document machine usage although other information is available in their Web home page.