Ongoing Research
The objective of the Unify project is to develop a scalable distributed shared memory multicomputer from the power of workstations interconnected with high-speed and wide-area networks by employing a hierarchical memory pace, scalable synchronization primitives, and flexible shared-memory consistency model.
MBFS is a distributed file system based on the use of volatile memory storage that forms a massive distributed cache decreasing both read and write latencies. MBFS introduces a new file system abstraction allowing among other things persistence specification, quality of service specification, and varying degrees of consistency.
The TMTP Project is investigating a reliable Tree-based Multicast Transport Protocol.
The objective ofActiveCast Project is to develop a set of services for Active Networks.
There are currently several ongoing research projects to develop instrumentation systems and tools for performance monitoring and modeling of parallel and distributed systems.
Techniques are being developed for checkpointing and roll-back for large scale distributed applications.
There are currently two projects in the area of multimedia. The Electronic Beowulf project is studying issues of the digitization, representation, archival and access of library manuscripts and artifacts. The MOODS Project (Modeling Object-Oriented Data Semantics) is investigating how one can effeciently identify and represent content in multimedia data, regardless of the application domain.
We have several projects focusing on developing algorithms that are able to perform traditional image processing tasks with compressed image data. The ability to operate directly with compressed data saves time and bandwidth, eliminating the need for full decompression. Because compression is central to the problem of transferring video streams in real-time from a sensor across a network for processing, having algorithms that can exploit the compressed data is a logical step in real-time applications. We are currently focusing on applications such as principle component analysis and stereo vision.