Managing trade-offs between information loss and latency in real-time distributed systems

Abstract

Multicasting is used in many real-time distributed applications such as market data feeds, videoconferencing, interactive distance learning. The need for multicasting is either because the databases are replicated or simply because there are multiple destination nodes involved in the communication process. However consistency problems are inherent in distributed environments where the variable nature of inter-communication delays between sites may interfere with the correct order of the received data. Several consistency preserving algorithms rely on resequencing methods. In this paper, we use partial order resequencing in order to meet the temporal constraints of these real-time systems. This may cause loss of data and could affect the data integrity. By using simulation results and approximate analysis, we analyze the trade-offs between minimizing the latency and keeping the information loss acceptable.