Guaranteeing and Explaining Stability across Heterogeneous Load Balancing using Calculus Network Dynamics

TL;DR

This paper introduces a calculus-based framework to analyze and guarantee the stability of load balancing across heterogeneous base stations, addressing oscillations and providing theoretical synchronization conditions.

Contribution

It develops a novel calculus network dynamics model that explains load oscillations and offers stability guarantees for diverse network topologies.

Findings

Established synchronization conditions for load balancing dynamics.

Designed mechanisms to mitigate oscillations and ensure convergence.

Provided a theoretical foundation linking network topology and load stability.

Abstract

Load balancing between base stations (BSs) allows BS capacity to be efficiently utilised and avoid outages. Currently, data-driven mechanisms strive to balance inter-BS load and reduce unnecessary handovers. The challenge is that over a large number of BSs, networks observe an oscillatory effect of load evolution that causes high inter-BS messaging. Without a calculus function that integrates network topology to describe the evolution of load states, current data-driven algorithms cannot explain the oscillation phenomenon observed in load states, nor can they provide theoretical guarantees on the stability of the ideal synchronised state. Whilst we know load state oscillation is coupled with the load balancing process algorithms and the topology structure of inter-BS boundary relations, we do not have a theoretical framework to prove this and a pathway to improving load balancing algorithms. Here, we abstract generic and heterogeneous data-driven algorithms into a calculus dynamics space, so that we can establish the synchronization conditions for networked load balancing dynamics with any network topology. By incorporating what is known as "non-conservative error" and the eigenvalue spectrum of the networked dynamics, we can adjust the inter-BS load balancing mechanisms to achieve high efficiency and convergence guarantee, or to mitigate the oscillation when the synchronisation condition cannot be satisfied.