Data-Driven Internet Weather Maps: Real-Time Analysis of Global Network Performance Fluctuations

Abstract

In this paper, a data-driven model of creating Internet weather maps is introduced to visualise current network performance around the world. The system analyses the latency, jitter and packet-loss data using passive telemetry on more than 500 distributed monitoring nodes and compiles them into a scalable multi-layered visualisation model. The raw telemetry is converted to colour mapped maps of network climate zones around the globe by advanced interpolation and clustering algorithms. The suggested system is based on the distributed edge-cloud processing and GPU-accelerated rendering to provide responsiveness and scalability. A twelve-month analysis indicated the presence of apparent spatio-temporal variations in Internet performance in form of congestion propagation, cross-continental asymmetry and routing instability. The findings indicate that Internet weather maps can be used as a visual decision-support tool by the operators and researchers who follow the trends in Internet resilience and performance across the world.