Multi criteria decision system for distribution provisioning and resource optimization using analytical hierarchy process

摘要

The docker container is an operating-system-level virtualization that enable users to run isolated linux systems (containers) on a host. It takes one container on an existing host server and each container will use the resources on the host server for each isolated service or application. As long as the container is active, it takes the resource of the host server and is highly possible to continuously grow. This will cause problems due to the limited resources. Multiple servers then are employed to provide an ever-expanding app or service. However, when using multiple servers or hosts to run the service, resource availability on each server often varies. Hence, the distribution tasks over container without considering the resource availability in each server may lead to inefficiency. In this work, we use well-known Multi Criteria Decision Making (MCDM) methods, namely Analytical Hierarchy Process (AHP) to determine which server is fit to serve the request when container provision comes. AHP utilizes the information of the availability of memory, CPU, and file storage as parameters in performing this task. Moreover, the first two parameters along with LTA (Last Time Accessed) information are further utilized by AHP to decide which server to be shut down.