Remaining Capacity Based Load Balancing Architecture for Heterogeneous Web Server System

Load balancing system can improve the performance of a popular Web server. Most of the load balancing systems use homogeneous Web servers as their backend servers. In this situation, the dispatching algorithm can obtain load sharing among the Web servers quite easily. However, when the Web server system contains servers with different capacities, we must use factors such as the remaining capacity of these Web servers as the criteria for decision making. In our proposed system, we use the remaining capacity as the parameter of the dispatching algorithm and compare the performance with systems using round robin and least connection algorithms. The hit rate, utilization, and drop rate are compared. We also derived a formula to calculate how many servers are needed when the system is overloaded and what kind of capacity the server is required.