CAUSAL DELIVERY OF MESSAGES WITH REAL-TIME DATA IN UNRELIABLE NETWORKS

Causal order states that for any process the order in which it is deli vered messages cannot violate the happened-before relation of the corr esponding sendings. Such a communication abstraction has been defined for reliable distributed systems in which data of application messages have unlimited time validity. In this paper we extend the notion of c ausal order to cope with unreliable communication networks in which me ssages have real-time delivery constraints. In particular, we assume t hat messages have a limited rime validity, a, after which their data c an no longer be used by the application, and that some of them can be lost by the communication network. This new abstraction, called Delta- causal order, requires to deliver as many messages as possible within their validity time in such a way that these deliveries respect causal order. Two efficient implementations are proposed in the case of one- to-one and broadcast communication. Examples of distributed multimedia real-time applications, in which scheduling messages deliveries respe cting a-causal order is a crucial point for the quality of the service , are given.