eCube: Hypercube Event for Efficient Filtering in Content-Based Routing

Abstract

Future network environments will be pervasive and distributed over a multitude of devices that are dynamically networked. The data collected by pervasive devices (e.g. traffic data, CO2 values) provide important information for applications that use such contexts actively. Future applications of this type will form a grid over the Internet to offer various services and such a grid requires more selective and precise data dissemination mechanisms based on the content of data. Thus, a smart data/event structure is important. This paper introduces a novel event representation structure, called eCube, for efficient indexing, filtering and matching events. We show experimental results that demonstrate the powerful multidimensional structure and applicability of eCube over an event broker grid formed in peer-to-peer networks.

Distributed spectral cluster management

“…The second approach (denoted as OV ) is a representative of many prominent publish/subscribe systems [7,31]. It considers the absolute overlap between the subscriptions as a measure of their similarity and uses k-means or R-tree [39] algorithms to cluster them.…”

“…Both variants use a two-level hierarchical organization to perform dimensionality reduction and calculate subscription clusters. The OV-RTree approach implements a distributed R-tree similar to [7,39]. Figure 3(k) shows the notification cost during the evolution of the system as more and more events are disseminated.…”