High number of users, connected devices and services on the Internet producing high traffic and load on the web servers causes a degradation of the quality of Internet services. A possible solution to this problem is to use a cluster of web servers. The cluster requires a load balancer to provide scalability and high performance of the services offered. The main load balancer is the only entry point to the server cluster in this architecture. In this paper, the researchers propose a two-tier load balancer rather than a single one to achieve more scalability and reduce the load on the main load balancer. The study also compared three features, Response Time Average, the load balancer CPU Utilization, and Servers’ CPU Utilization. The comparison uses three algorithms (Round Robin, Number of Connections, and Least Load) through two experiments. The results concluded that the Multi-Tier Load Balancing method offered better CPU utilization than a Single-Tier Load Balancing method for Round Robin and Server Load algorithms. However, the Single-Tier Load Balancing method provided better response time for all three algorithms. Moreover, the Round Robin and Server Load algorithms using a Multi-Tier method balanced the CPU utilization for all servers. This result shows the Multi-Tier method handles huge traffic and large number of servers with better CPU utilization.
Key words: Round Robin; Least connection; server load, Load balancer; Multi-tier load balancer; Server Cluster.
|
