Forecasting Traffic and Balancing Load for Quality Driven LTE Networks

[thesis] [pdf] [cite]

Abstract

With the current increase in network traffic in radio networks, it is now more important than ever to manage this traffic efficiently in order to utilise the available network resources intelligently. The goal of this thesis is to provide means for the operators to optimise their networks regarding the management of load across the network.

This thesis proposes a two part approach to load management, an autonomous load balancing algorithm in conjunction with a traffic forecasting methodology.

The proposed load balancing algorithm works in closed loop where each cell measures its load and its neighbors load in order to adjust its handover parameters to offload traffic to other cells. Several simulations were run in order to validate the concept, the utilisation of this method decreased the average number of unsatisfied users in a network up to 4%, depending on the network configuration. The simulations were run using the MATLAB Vienna LTE System Level Simulator, which was heavily modified for this work.

The forecasting tools and methodologies discussed in this work were mostly taken from the field of economics and were shown to work extremely well when used to forecast network traffic, whether it be data or voice. Some of the proposed techniques were shown to predict network traffic two months in advance with a median error across 86 cells of just 14%.

This approach shows the potential of reducing the amount of wasted network resources and increase savings for the operator.