[Home] [People] [Papers] [Software] [Research]

The Transis Project Research Efforts



To be updated...

PrFQ (Probabilstic Fair Queueing) is a new probabilistic approach for packet fair scheduling.

Support for CSCW and Multi-media Applications

An important objective of our research is exploting (and extending) Transis for a framework for Computer Supported Cooperative Work (CSCW) applications, such as multi-media and desktop conferencing. Collaborative computing may greatly benefit from the group communication paradigm, and also require QoS communication. The concept of the Multi-media Multicast Transport Service (MMTS) incorporates QoS communication into the group communication framework. We are focusing on efficient communication solutions for a variety of QoS requirements, for example ATM QoS options (see above) and efficient totally ordered multicast.

The scalable fault tolerant Video-on-Demand Service exploits Transis for achieving the fault tolerance, and the MMTS attitude for incorporating QoS. Other collaborative applications that exploit Transis include a shared white-board, a distributed MIDI jam session, and a distibuted window manager that allows sharing of X windows.

Caelum is a comprehensive toolkit for the development of highly available groupware and CSCW applications, using the MMTS as a building block; it provides efficient tools for maintaining consistency of distributed and replicated information in the face of faults. Examples of such services are state transfer, object replication (the porotocls of COReL and robust replication), a dynamic voting scheme for primary components and atomic commit.

Collaboration with the Horus Group at Cornell University

TransisE unifies Transis and Ensemble, it implements the Transis LAN reliable messaging subsystem in the Ensemble framework. A new membership service for supproting CSCW applications, is currently developed over both Transis and Ensemble, using the Maestro system.

Secure Group Communication

A generic security architecture was added to the Ensemble system. The architecture includes several protocols, and interfaces to standard security packages and authentication services. The system can create secure groups such that: We have investigated key management issues, and fast rekeying in the presence of untrusted members.

Using Group Communication for Distributed System and Network Management

The Transis-based network management toolkit provides solutions for the following important classes of management tasks: maintaining consistent distributed configurations (improvement of NIS), software distribution to a large number of destinations and simultaneous command execution on a number of machines. This toolkit is being extended to an SNMP based network management toolkit, Snmp Groupware (SnmpG).

Formal Specifications of a Group Communication System

It comes true that even when two people talk about such a well-known term as Virtual Synchrony, they do not always mean the same thing. So, what do we exactly provide and what do we assume on the underlying network ? How applications over group communication could benefit from these properties ? How much do we pay to implement them ? Is it possible to specify the agreement on a group membership weak enough to be implementable in an asynchronous prone to failures environment but strong enough to be useful for applications ? Are failure detectors of any aid in building specifications ? How to model the "best effort" requirement ?

The former efforts of specifying particular services of a group communication system are Extended Virtual Synchrony definition and A Framework for Partitionable Membership Service. The applications that exploit these services are described in state transfer and object replication works (the protocols of COReL and robust replication). More effort is put now in order to collect the pieces together, to provide more strict and correct specifications, to explain and illustrate things that have not found a proper expression in related work till now.

Last modified: Thu Apr 24 20:39:10 IDT 1997