David Farber - US grants

This award will provide infrastructure for research that is organized around five laboratories: 1. LINC - for research on artificial intelligence and natural language processing; 2. GRASP - for research on machine perception and robotics; 3. GRAPHICS - for research on graphic interfaces, movement description, and animation; 4. DSL - for research in computer architecture and computer communication; 5. LOGIC & COMPUTATION - for research in logic and computation, including theory of computation, database systems, and programming languages. Two new facets of the research, integration and upward scaling, require an enhanced experimental environment involving machines with massively parallel architectures. The award will help to develop this environment by providing funds for a SIMD machine for work in natural language processing, and active perception and real time manipulation; a MIMD machine for simulation and research involving extensive scientific calculations; as well as high speed workstations with rich environments for work in theoretical computer science.

This award is for the acquisition of teleconferencing and multimedia technology, and for curriculum changes to expand a curriculum in Telecommunications. The University of Pennsylvania will use a new instructional delivery system, the "video wall", to develop "Telementoring" as a long distance learning technique. The video wall is already being used in research projects by members of the AURORA Gigabit Testbed which is being supported by both NSF, DARPA and a consortium of industrial research partners. Educational materials developed will be made available to other academic institutions through Internet, and results of the educational experiments will be disseminated through publications and presentations at educational and professional meetings. The University of Pennsylvania plans to use a state of the art instructional delivery system, the "video wall", to provide multimedia and teleconferencing support for undergraduate courses in telecommunications. The video wall is an experimental video conferencing terminal with two large screen projection televisions mounted side-by-side creating the illusion of one large screen. Two cameras, co-located with the screens, are arranged to produce a single blended life- size image which is combined with high-quality directional sound. The results of the curriculum and materials development will be disseminated through Internet, and publications and presentations at educational and professional meetings.//

CDA-9703220 Bajcsy, Ruzena University of Pennsylvania Asymmetric Bandwidth Channels: Applications to Real-Time Computing and Robotics This award is for the acquisition of infrastructure to support research which is to investigate a cost-effective and broadly deployed communication model, Asymmetric Bandwidth Channels, for which few abstractions in computer science are available. Research into the application of systems characterized by low-bandwidth interactive channels between clients and server, and a high-bandwidth broadcast from server to clients will be conducted. The proposed work will develop communications abstractions that applications can effectively use for such an infrastructure, and computational models for these abstractions. Model performance will then be evaluated on a testbed of multiple semi-autonomous robotic agents. The challenging problems to be addressed in this project include: (1) selecting which path to take from server to client; (2) determining the degree of broadcast channel sharing possible in a computer communications environment; and (3) scheduling transmissions from the shared broadcast terminal. The research will also target three fundamental problems underlying coordination of robotic agents: (1) development of world models based on observations of individual agents and exploration of an unknown or a partially known environment to build a complete model; (2) task planning based on the world model while accounting for possible uncertainties and latencies; and (3) control of robotic agents based on visual and other sensory information.

The fundamental project hypothesis is that management and control of the Internet is the ideal
application of active networks; the researchers propose a system architecture to test this.
Active Networks are constructed from elements, such as packet routers, allowing programmability
on a per-user or even per-packet basis. With the new software capabilities available from systems such as Caml and Java, active networks offer the promise of more rapid adaptation to changes in technology or requirements, and more rapid introduction of new services. These potential advantages come with the disadvantages of increased complexity, and its consequences for performance and security.
Early prototype systems (ANTS, CANES, Smart Packets, SwitchWare and others) illustrated
various points in the design space, trading off among usability, performance, and security. The
prototypes demonstrated first, that such systems could be built, that applications did indeed exist,
(e.g., Active Bridging and Active Reliable Multicast), and second, that they performed well enough
(10-100 Mbps) to handle the throughputs of almost all current Internet access points. Thus much
of the "edge" of the Internet can add active network capabilities with minimal performance impact.
A more interesting possibility exists, that of using active networking technology to incrementally
activate the IP Internet. The researchers believe this can be achieved, as described within the proposal, by
co-locating programmable elements with IP routers capable of fast packet forwarding. The researchers have
experimented with this idea on a small scale and it offers considerable promise for increasing the
manageability of the Internet with its exponential increases in scale.
The Global Active IP Network (GAIN) project represents the University of Pennsylvania's
research program as part of a larger 10M Euro research effort (FAIN). FAIN was considered and
top-ranked within the E.U. IST Programme competition. European members were funded, with the
expectation that Penn would seek funding from U.S. sources. The consortium includes University
College London (UK), the Jozef Stefan Institute (Slovenia), the National Technical University of
Athens (Greece), the Universitat Politecnica de Catalunya (Spain), Deutsche Telekom Berkom
(Germany), France Telecom/CNET (France), KPN (Netherlands), Hitachi Europe Ltd. (UK),
Hitachi Ltd. (Japan), SAG ICN (Germany), ETH Zurich (Switzerland), GMD Forschungszentrum
Informationstechnik (Germany), IKV++ (Germany), INTEGRASys (Spain), and U. Penn in the
United States. This proposal to NSF is a request for funds to support Penn in this international
consortium.
Penn's focus with GAIN is applications of Active Networks to IP network resource management
and security. The researchers will investigate the prevention and mitigation of sophisticated "denial of service" attacks on security. The researchers are playing a strong role in experiment definition and evaluation for FAIN. This proposal to NSF provides background on Active Networking, outlines the research goals for an active IP network, sets this work with the context of FAIN, and argues the importance of providing U.S. participation in a truly global consortium with European and Japanese collaborators.
( The FAIN proposal has been provided to NSF.)

The goal of this research is to remove the tight binding of user
state to specific computing hardware. This tight binding arose
at the birth of personal computing nearly three decades ago, and
has continued unchanged since that time. It has many negative
consequences: hardware upgrades are painful, disaster recovery is
complex, and total cost of ownership is high. It also forces
mobile computing into a paradigm of each user carrying personal
hardware, rather than traveling hands-free and taking advantage
of pervasive hardware.

The proposed research aims for a new vision of on-demand personal
computing called "Internet Suspend/Resume (ISR)" that exploits
the ubiquitous presence of the Internet. This research explores
issues at the intersection of two well-established technologies:
virtual machine (VM) technology and distributed file system
technology. By layering a VM on a distributed file system that
performs aggressive client caching, a user's entire personal
computing environment (operating system, applications and user
files) can be accurately and safely delivered anywhere on the
Internet.

The proposed research spans operating systems, distributed
systems, mobile and pervasive computing, and security and
privacy. Its has three major thrusts: (a) Coping with Huge VM
State, (b) Preserving Privacy of VM State, and (c) Dynamically Varying
Client Thickness. The work in each thrust includes conceptual
development of novel techniques and algorithms, as well as their
implementation and experimental validation in the context of a
prototype. This prototype will be disseminated in open source
form, for use by other researchers and by industry.