Control mechanisms of presence updates: A tradeoff between traffic optimization and information consistency

Abstract

Presence information, which consists of a wide variety of user context, is seen as a key enabler of user personalization in multimedia applications. Presence protocols tend to generate much traffic due to timely dissemination of presence changes and presence subscription maintenance. On one hand, controlling the rate of presence publications saves presence traffic on the access network link. However, low publication rates may introduce excessive delays in presence applications, which usually have real-time constraints. We therefore propose a strategy for adjusting publication rates to the frequency of presence changes based on the Markov theory, which are referred to as sojourn-based rates. On the other hand, presence overload is specially challenging in large-scale presence federation scenarios, in which there are millions of subscriptions between the federated domains. We propose a queuing system that controls the rate of both presence notifications and presence publications in these scenarios. We also propose an adaptive control mechanism that dynamically changes the queuing system’s parameters in order to satisfy the federated domains’ Quality of Service (QoS) needs on traffic optimization and information consistency. This mechanism can be combined with sojourn-based rates in order to improve its performance. We provide a mathematical model of both proposals and analytically estimate their performance.