Real-Time Communication in Quasi-Synchronous Systems. Providing Support for Dynamic Real-Time Applications

Carlos Almeida and Paulo VerĂ­ssimo

INESC Technical Report

Abstract

This paper addresses the problem of having real-time group communication in systems that are not completely synchronous (something we refer to as quasi-synchronous). This situation can happen when load is not completely controlled thus not ruling out overload scenarios. This is usually related to dynamic characteristics of applications and/or environment that make the use of a resource adequacy policy not cost-effective. In such systems, temporal guarantees can only be made to the highest priority messages or small bandwidth specific communication channels.

In our model, we use the highest priority messages or a small bandwidth synchronous channel to implement a timing failure detection service. This service is used as an oracle by the group communication protocols that are thus able to provide safety in a timely fashion. Having a timely detection of timing faults, allows the application to dynamically and safely adapt to a changing environment. It is thus possible, for the applications, to define several different working envelopes corresponding to several different qualities-of-service, and evolve, in a controlled way, between those envelopes.

Also available extended report (gzip postscript).