Laboratory for Networking and Distributed Computing
|
Mission Facilities Questions Projects |
|
Mission
The goal of this laboratory is to promote experimental systems research in the areas of networking and distributed computing. In order to support this research we must develop and maintain an experimental computing facility. Fortunately, the laboratory has been seeded with sufficient equipment to launch a small number of initial investigations and we expect that results from these early investigations will provide the leverage needed to grow the lab. Although the purpose of the lab is to conduct fundamental systems research, many of our efforts will be motivated by the need to support one or more real world applications. In some cases members of the lab will directly engage in applications work to gain a deeper understanding of the underlying systems issues. A critical component in these early stages of the lab is the development of sufficient systems software infrastructure to support the initial research. To this end a number of the early projects involve the systems software implementations.
Two representative areas of research interest are:
- The design and analysis of algorithms for call admission with quality-of-service guarantees in ATM networks.
- Runtime support for distributed shared memory applications.
- Implementation of a multithreaded ATM driver for BeOS
- Implementation of UNI signalling
- Runtime support for distributed BeOS applications
-
- Video server scheduling
- Medical image processing;
- Earthquake fault analysis.
Fundamental Questions
Facilities
-
Currently the lab consists of:
-
1 FORE 8-port ATM switch
1 GTE 4-port ATM switch
2 dual PowerPC 603e BeBoxes, BeOS
2 120MHz Pentiums, Windows NT
1 SGI Indigo2 Workstation, IRIX 5.3
2 120MHz Pentiums, Windows NT
1 SGI Indigo2 Workstation, IRIX 5.3
- Our plans for upgrading the lab:
-
Quality of Service Network Architectures
This work involves: (1) the design and implementation of traffic control mechanisms for dynamic resource allocation, application-specific networking support, and network-aware adaptive applications; (2) the use of traffic measurements to guide the design and implementation of computer-aided performance prediction tools.
Medical Image Processing
This work focuses on the development of parallel and distributed algorithms for the analysis of various types of medical images (MRI, x-ray, ultrasound). Currently we are experimenting with a number of different transform techniques (FFT, WHT, DCT and Number-Theoretic Transforms) for processing breast cancer x-ray images.
Algebraic Image Processing
Freshman Research Project
Mapping of Scientific Computations
The objective of this project is to develop effective algorithms for mapping scientific computations defined on regular topologies (mesh and hypercube) to networks of workstations (NoWs) of arbitrary topology. Of particular interest is the use of ATM to minimize latency.
Parallel Scientific Computations
This work focuses on parallel algorithms for earthquake fault analysis, crystallographic and quasicrystallographic computations, group and graph computations, large-scale combinatorial searches, as well as wavelets and joint time frequency analysis. New visualizations for these computations are being developed.
Hybrid Architectures for Parallel/Distributed Computations
This project concerns the development and deployment of high performance computing engines as add-on boards for standard PCs or workstations. in a distributed shared memory environment.
Simulation of Parallel I/O Architectures
Visualization of Large Scale Searches
K-12 Multimedia Applications
K-12 Multimedia applications have dual value: (1) they allow the development of applications which can potentially be used in a K-12 classroom setting and the investigation of all the related software development issues; (2) they can provide a workload for the QoS work described above.
| Web page maintained by Dennis A. Dean II | Web page update 5/18/98 |