LightningNet: Distributed Graph-based Cellular Network Performance Forecasting for the Edge

TL;DR

LightningNet is proposed, a lightweight and distributed graph-based framework for forecasting cellular network performance, which can capture spatio-temporal dependencies that arise in the network traffic.

Abstract

The cellular network plays a pivotal role in providing Internet access, since it is the only global-scale infrastructure with ubiquitous mobility support. To manage and maintain large-scale networks, mobile network operators require timely information, or even accurate performance forecasts. In this paper, we propose LightningNet, a lightweight and distributed graph-based framework for forecasting cellular network performance, which can capture spatio-temporal dependencies that arise in the network traffic. LightningNet achieves a steady performance increase over state-of-the-art forecasting techniques, while maintaining a similar resource usage profile. Our architecture ideology also excels in the respect that it is specifically designed to support IoT and edge devices, giving us an even greater step ahead of the current state-of-the-art, as indicated by our performance experiments with NVIDIA Jetson.

Figures (9)

• Figure 1: High-level architecture of the measurement and forecasting infrastructure integrated in the cellular network.
• Figure 2: LightningNet Architecture. The design of the sub-classifier (SC) is depicted on the uppermost part of the diagram, whilst the lowermost part depicts the Hierarchical Model (HM) structure.
• Figure 3: Graph partition into $k$ equally sized sub-graphs. The sub-graphs with the highest concentration are used to train the LightningNet model, which then can be applied on any given sub-graph in the system.
• Figure 4: System design flow of LightningNet. Each antenna in the cellular network provides the hourly collected KPIs to the cloud server. Through message passing, neighbouring antennas are sharing information and the final embeddings are calculated through an aggregation function from the features of all neighbours.
• Figure 5: Average Precision (a,b,c) and Recall (d,e,f) for SG$_{1-7}$, forecasting horizon hz$\in \{12,24,48\}$ and memory buffer mb$\in \{12,24,48\}$ hrs.