Most web servers that are used today still use single backend server architectures. The problem that arises is how the single server is able to handle large data requests. We must consider using web server clustering to improve web server reliability. We built this cluster using virtualization technology such as virtual containers. One of the container-based virtualisations currently is Docker. However, managing multiple containers to make a single service is a challenging task. Docker introduces a distributed system development tool called Docker Swarm. We propose a load balancing mechanism on Docker Swarm to balance internal load, so that it can distribute requests to web server. In addition, we also try to find out the performance of roundrobin and leastconn algorithms on loadbalancing mechanisms that will be used on Docker Swarm. The test results show that loadbalancing applying the least connection algorithm has a throughput of 15 Mbps in 1000 requests, 17 Mbps in 3000 requests, 17 Mbps in 5000 requests, while the round robin algorithm has a throughput of 15 Mbps in 1000 requests, 14 Mbps in 3000 requests, 15 Mbps in 5000 requests. The results show the least connection algorithm has better performance than the round robin algorithm. In addition, the results of data distribution are balanced on each available web server.
Copyrights © 2019