Dynamic Redundant Query Forwarding In Mini- Datacenter: A Bio-Inspired Approach

In this paper, we present a novel concept of modeling wireless sensor net-work (WSN) as a mini-datacenter to facilitate IoT data collection, thus reducing the dependency on centralized cloud resources. The mini-datacenter is comprised of a command center, a set of gateways and sensors, where each gateway is equipped with limited buffer capacity. The incoming traffic ar-rives in bursts of varying intensity, which may cause buffer saturation and data loss. To address this, a dynamic redundancy computing mechanism is implemented, adapting the duplication of queries based on real-time traffic intensity. However, selecting a set of redundant gateways randomly from the available pool according to the computed redundancy level often fails to bal-ance the load distribution, leading to saturated buffer spaces in certain nodes. To address this limitation, we have formulated the redundancy-buffer occupancy problem as an optimization problem with an objective function to dynamically adjust redundancy levels while ensuring balanced load distribution, based on real-time buffer occupancy and the characteristics of incoming traffic. We employ a bio-inspired evolutionary algorithm, specifically a Genetic Algorithm (GA) to intelligently select the distribution weights of gateways by considering real-time network conditions. Experimental evaluations conducted using the proposed mini-datacenter setup with 10 gateways and a set of connected sensors demonstrated that GA-based query-forwarding outperformed the rule-based query-forwarding by achieving an 85% reduction in the buffer saturation events, and a 13.74% reduction in average buffer space occupancies.