The Comparison of CPU Utilizations In The Different Load Balancer Scheduling Algorithms
This is an academic report for the project of the CENG 530 Computer Networks and Communications course. In the project, CPU utilizations of random, round robin and distribution according to the server load scheduling algorithms for load balancers are studied.
The purpose of this experiment is to observe the performance of a load balancer in a network environment by using different scheduling algorithms via simulations. The simulations are created by Riverbed Modeller Academic Edition 17.5. Four main scenarios are constructed for the experiment simulations; a system without using load balancer, three systems using load balancer with randomly choosing, round robin and server load algorithms respectively.
Load balancer definition…
The identical client machines in the system are configured as (…..). Identical servers used both on scenarios with and without load balancer are configured as (…..). (load balancer configuration) The performance of the systems are evaluated by utilizations of CPUs per servers. Equally distributed average CPU utilization per servers are considered as ideal situation.
In the beginning, a system without load balancing is simulated as the control group. As there is no load balancer, only one server handled all the client requests. The CPU utilization values are given in Table 1.
The number of the identical servers are increased from two to eight at each experiment.
The random distribution scheduling algorithm is used as the second main scenario. In the random distribution scheduling algorithm provided by Riverbed Modeller, load balancer selects servers randomly to handle the client requests. Below the results of CPU usage for two servers are presented:
When compared to the control group, the average CPU utilization values of the servers in the system with load balancing are decreased by more than half of the previous values. As the number of the servers are increased to eight, average values are decreased more dramatically as given below: