Dynamic Approaches to In-Network Aggregation

TL;DR

This paper introduces extensions to distributed aggregation protocols that enable mobile wireless devices to accurately estimate global properties like count, sum, and average despite high mobility and dynamic network conditions.

Contribution

It presents novel modifications to existing unstructured aggregation protocols to maintain accurate estimates in highly dynamic, mobile wireless environments.

Findings

Protocols effectively estimate aggregates despite node mobility.

Extensions handle unexpected peer departures and arrivals.

Analysis confirms robustness in dynamic environments.

Abstract

Collaboration between small-scale wireless devices hinges on their ability to infer properties shared across multiple nearby nodes. Wireless-enabled mobile devices in particular create a highly dynamic environment not conducive to distributed reasoning about such global properties. This paper addresses a specific instance of this problem: distributed aggregation. We present extensions to existing unstructured aggregation protocols that enable estimation of count, sum, and average aggregates in highly dynamic environments. With the modified protocols, devices with only limited connectivity can maintain estimates of the aggregate, despite \textit{unexpected} peer departures and arrivals. Our analysis of these aggregate maintenance extensions demonstrates their effectiveness in unstructured environments despite high levels of node mobility.