Load balancing in modern network infrastructures — A simulation model

This paper presents a critical review of load balancing policies of network traffic loads composed of TCP/IP application protocols and applications and discusses the performance improvements sustained by such policies. The review introduces principles of network design of cost efficient networking infrastructures in an economy of scale suitable for global internet network traffic loads. Discrete event simulation modeling of a linear network architecture is used to study the effects of five network load balancing policies (Server Load, Round Robin, Number of Connections, Random, and Weighted Fair Queuing) against a baseline scenario of no-load balancing (NLB). The simulation study presented in this paper, clearly demonstrates that load balancing can lead to a reduction of application response times, improved CPU utilization for Server performance, and higher application throughput, thus improving network availability to the end user and the overall network resiliency.