Abstract :
The increasing number of requests to the web server causes the inability of server to service all requests from users. The solution to this problem is to implement load balancing. Load balancing has the ability to distribute requests from users across multiple paths in a balanced. However, in general load balancer server is built using physical devices which cause their use to be inef icient due to the need to change or add servers when the server cannot accommodate the increasing number of requests.. Cloud computing technology comes with the ability to add devices logically if administrator wants to improve server performance. This research aims to obtain the performance value of load balancing on a web server using the least connection algorithm on IaaS service. IaaS cloud is built on a local server using Openstack software. The test was carried out through 5 variations in the number of connections, 200 connections, 500 connections, 1000 connections, 2000 connections, and 5000 connections with a rate of 100 requests/second. The results show that the average throughput value of all variations is 1.580 Mbit/s with the highest value at 1000 connections, that is 1.859 Mbit/s. The delay indicates the "Very Good" category based on the TIPHON standard, which is less than 150 ms. The presentation of packet loss at 200 connections to 1000 connections shows the Very Good category, then at 2000 connections and 5000 connections respectively it shows the Good and Fair categories. The highest percentage of CPU usage was at 5000 connections at 62,565%. The test results show that the load balancing on the web server that is applied to IaaS services is still optimal if it is given a load of up to 5000 connections. Keywords: Web server, load balancing, least connection, cloud computing