Towards Dynamic Energy and Performance Management of Wireless Internet Service Provider Networks
Abstract
Wireless Internet Service Providers (WISPs) play a crucial role in bridging the digital divide, connecting millions in remote and rural communities worldwide where traditional wireline infrastructure is economically unfeasible. These networks operate under unique constraints: limited unlicensed wireless spectrum, dependence on renewable energy sources, and inadequate performance monitoring and management tools. This thesis presents an integrated framework of three complementary contributions to address these fundamental challenges. The monitoring and management gap is first tackled through Polus, a framework that detects and characterises latency under load from the subscriber perspective, while actively preventing bufferbloat through dynamic control of the CAKE qdisc. While conventional ICMP-based monitoring can overlook critical performance issues, Polus reveals how bufferbloat significantly degrades application quality in production WISP networks. By enabling precise congestion localisation without synthetic load generation, Polus provides operators with real-time visibility into the actual user experience, enabling proactive intervention before perceivable quality of service degradation occurs. Building on these monitoring capabilities, ElasticWISP-NG is introduced as a dynamic resource management scheme enabling energy-proportional operation of wireless access networks. Through its novel Energy State Outlook (ESO) metric and Enhanced Modified Best-Fit Decreasing (EMBFD) bin-packing algorithm, the system intelligently allocates network resources based on actual demand patterns and renewable energy availability. Evaluation with real-world WISP data demonstrates substantial energy savings of 64% (930.00 Wh daily) without compromising service quality, with some access points operating just 1.75 hours daily during peak demand periods. To complete this integrated approach, ElasticRoute is introduced – a practical traffic engineering solution that enables dynamic flow routing without requiring expensive, specialised hardware. By leveraging the mature Linux networking stack and eXpress Data Path (XDP) framework, ElasticRoute delivers Unequal-Cost Multipath (UCMP) capabilities on commodity hardware, increasing maximum achievable flow by 49% compared to traditional shortest-path routing in the proof-of-concept implementation. Unlike traditional Software-Defined Networking (SDN), ElasticRoute can operate gracefully during controller failure, making it suitable for dynamic wireless networks that lack a reliable control plane. Together, these three interdependent contributions form a comprehensive framework addressing the core technical barriers to sustainable WISP operations. The research demonstrates that with appropriate monitoring and control (Polus), energy management (ElasticWISP-NG), and traffic engineering (ElasticRoute), WISPs can deliver reliable, high-quality connectivity while significantly reducing operational costs and energy consumption. These innovations enable WISPs to extend their reach with limited resources, accelerating progress towards equitable Internet access for the world’s hardest-to-reach populations.
Fields of Study & Subjects
Wireless Networks and Protocols
Network Traffic and Congestion Control
Advanced Wireless Network Optimization
Publication details
- Publication date
- 16 February 2026
- Language
- EN
- Document type
- Thesis