<![CDATA[A growing number of internet users leads to a high amount of network traffic. Servers need to be added to avoid overload. Load balancing is also needed to distribute the load equally. Load balancing is widely implemented in SDN because it's programmable and flexible in a complex network. Therefore, this research proposes the implementation of load balancing in SDN using linear regression. The dataset for modeling linear regression function contains thirty samples. It contains the information of central processing unit (CPU), random access memory (RAM) and number of requests. The linear regression calculation is according to the server's CPU and RAM. The result is a prediction of the number of requests from each server. A server with the lowest request is chosen by the controller to handle the request. This research compares between linear regression and round-robin. In traffic distribution testing using linear regression, server with higher specifications gets more distribution than a server with lower specifications. In the testing of 300 requests per second, linear regression's response time is between 169.3 to 353.4 milliseconds, and round-robin is between 339.2 to 1232.6 milliseconds. Linear regression's throughput is between 241.5 to 310.7 KB/s, and round-robin is between 129.4 to 179.8 KB/s. In CPU utilization testing, the highest CPU usage in linear regression is 75 percent, and round-robin is 98 percent.]]>
Copyrights © 2022
