A Rule‐Based Server Activation Strategy for Dynamic Load Balancing and Energy Efficiency in Data Centers

Abstract

The energy costs of the information services offered in data centers stand out as one of the most significant expense items. This study proposes an innovative load balance methodology that aims to reduce the energy consumption of load balancing systems serving in data centers through dynamic server management. While traditional load balancing servers have a structure that consumes energy continuously, the proposed methodology offers a system design that activates servers only when needed. This experimentally designed system consists of an administrator server and three application servers. The administrator server analyzes the instantaneous CPU utilization information from the application servers, evaluates the load trends, and activates or deactivates the servers according to the determined threshold values. At this stage, the decision to include or remove application servers from the system is made with a “rule‐based” decision system. In the study, applications are made with four different CPU utilization threshold values, 50%, 60%, 70%, and 80%, and the system's energy consumption is monitored. The findings show that the proposed methodology provides the highest efficiency at the highest threshold and has significant potential in terms of energy efficiency; thus, the proposed methodology can be a promising method for developing energy efficiency strategies in data centers.