Online Algorithms for Information Aggregation from Distributed and Correlated Sources

TL;DR

This paper develops online algorithms for aggregating information from multiple distributed, correlated sources, balancing communication costs and latency, with proven competitive performance and practical effectiveness.

Contribution

It extends the dynamic TCP ACK problem to multiple correlated sources, proposing simple threshold-based algorithms with optimal competitive ratios in distributed settings.

Findings

Algorithms match theoretical lower bounds.

Algorithms perform well in simulations.

Practical testbed results confirm effectiveness.

Abstract

There is a fundamental trade-off between the communication cost and latency in information aggregation. Aggregating multiple communication messages over time can alleviate overhead and improve energy efficiency on one hand, but inevitably incurs information delay on the other hand. In the presence of uncertain future inputs, this trade-off should be balanced in an online manner, which is studied by the classical dynamic TCP ACK problem for a single information source. In this paper, we extend dynamic TCP ACK problem to a general setting of collecting aggregate information from distributed and correlated information sources. In this model, distributed sources observe correlated events, whereas only a small number of reports are required from the sources. The sources make online decisions about their reporting operations in a distributed manner without prior knowledge of the local observations at others. Our problem captures a wide range of applications, such as in-situ sensing, anycast acknowledgement and distributed caching. We present simple threshold-based competitive distributed online algorithms under different settings of intercommunication. Our algorithms match the theoretical lower bounds in order of magnitude. We observe that our algorithms can produce satisfactory performance in simulations and practical testbed.