High quality desktop video conference based on agent systems
Şerban Catrava · 1996
The purpose of the research described in this dissertation is to build a model and a performance subsystem that allows the development of high quality video conference applications. The proposed model has a distributed agent system as its underlying framework, and permits the development of applications that adjust themselves in real time to quickly respond to a changing set of conditions and parameters. The agents that comprise the underlying framework cooperate and coordinate efficiently to maintain good quality of service during a video conference session. The performance subsystem is another important objective of this work. The performance subsystem acts as an application controller and is composed of quality evaluators, a compensation loop, and a set of quality enhancers. The set of quality enhancers is the most significant part of the performance subsystem. Quality enhancers play key roles in increasing and stabilizing the overall application performance. They analyze the system behavior and activate a recovery loop to compensate for the hardware or network overloads. The research focuses on quality enhancers that implement workload balancing and graceful degradation. Such quality enhancers enable a conference to continue, possibly at a lower pace and with compromised performance, even if some resources, hardware or software, are disabled. Allowances are made in the model to permit learning mechanisms to be added to the performance system. Learning mechanisms allow the agents to adapt in time and improve their behavior as more knowledge about the environment is acquired. This research provides the necessary tools to integrate distributed multimedia applications in a similar look-and-feel environment, and creates a powerful framework that allows high quality video conference systems to be developed efficiently and at reasonable costs.