Mochamad Rexa Mei Bella, Mahendra Data, Widhi Yahya
Recently, container-based virtualization is gaining popularity. It is a lightweight virtualization, which utilizes the capabilities of the Linux kernel to allow sandboxing processes from one another and controlling their resource allocations. One of the most used container-based virtualization today is Docker. Docker is an open source project designed for developing, shipping, and running an application inside a container-based virtualization. We can deploy several web application container using Docker to serve millions of users. It also reduces the possibility of a single point of failure in the architecture. However, managing several containers for creating a single service is a challenging task. Docker solves this problem by providing container cluster management called Docker Swarm. The Docker Swarm internal load balancing mechanism focused on how to distribute the request to the worker equally based on the user request. It does not provide any mechanism to monitor the resource utilization of each host machine. It is problematic because it can lead to unequal load distribution between the host machines. This research purpose is to distribute web server traffic inside a Docker swarm based on the resource utilization of the host machine. We focused on the memory utilization on each host machine. We propose the mechanism to monitor the memory utilization of each host machine and distribute the web traffic based on the memory utilization of each host machine. The result of our experiment is promising. Each of the worker nodes receives a fair distribution of load. This technique can decrease the possibility of a single point of failure in the web server cluster. © 2018 IEEE.
Faculty of Computer Science, Universitas Brawijaya, Malang, Indonesia
