Dimension-Based Subscription Pruning for Publish/Subscribe Systems

Abstract

Subscription pruning has been proven as valuable routing optimization for Boolean subscriptions in publish/ subscribe systems. It aims at optimizing subscriptions independently of each other and is thus applicable for all kinds of subscriptions regardless of their individual and collective structures. The original subscription pruning approach tries to optimize the event routing process based on the expected increase in network load. However, a closer look at pruning-based routing reveals its further applicability to optimizations in respect to other dimensions. In this paper, we introduce and investigate subscription pruning based on three dimensions of optimization: network load, memory usage, and system throughput. We present the algorithms to perform prunings based on these dimensions and discuss the results of a series of practical experiments. Our analysis reveals the advantages and disadvantages of the different dimensions of optimization and allows conclusions about the suitability of dimension-based pruning for different application requirements.