NETWORK LOAD BALANCING METHOD BASED ON FUZZY CLUSTERIZATION
DOI: 10.31673/2412-4338.2025.038726
Abstract
The current problem of ensuring effective load distribution in modern computer networks operating in conditions of rapid traffic growth, high dynamics of traffic profiles and variability of service quality requirements from users and applications is considered. In particular, it is emphasized that the spread of streaming video services, interactive AR/VR platforms, the Internet of Things, autonomous systems and distributed computing creates additional challenges to the infrastructure, which traditional approaches to load balancing are unable to effectively solve. Insufficient adaptability and strict decision-making rules in classical algorithms, such as round robin or threshold methods, lead to irrational use of resources and can cause overloading of individual nodes when there is free capacity in other parts of the network. The conceptual architecture of the system is presented, the main stages of its functioning are described - from collecting metrics and clustering to identifying candidate nodes for offloading and making decisions on traffic redirection. The nodes, which receive a variable load, was carried out. Three balancing approaches were compared: round robin, adaptive threshold algorithm (standard autoscaling with two load thresholds), and the proposed fuzzy cluster algorithm. The experimental model simulates the behavior of the system in a virtual network environment, which allows us to compare the effectiveness of the proposed method with traditional approaches. The analysis demonstrates improved load distribution, reduced probability of overloads, and increased system adaptability.
Conclusions are drawn regarding the prospects of implementing fuzzy algorithms in automatic network resource management systems. The feasibility of further research aimed at improving clustering methods, their combination with expert rules, and integration with intelligent agents within the framework of Software-defined Networking architectures is substantiated.
Keywords: computer networks, load balancing, fuzzy clustering, Fuzzy C-Means, adaptive algorithms, artificial intelligence.
