Fuzzy Intelligent System for Improving High Availability of a Docker-Based Hybrid Cloud Infrastructure

Abstract

Hybrid cloud infrastructure offers several benefits including high availability. However, one of the biggest challenges for the hybrid cloud infrastructure is to maintain the same high availability of the local or traditional infrastructure as this is normally maintained by the organisation. Docker Swarm facilitates the development of a hybrid cloud infrastructure based on Swarm nodes consisting a high availability feature, which guarantees the successful running of a Docker Swarm cluster for a certain number of failures depending on the number of primary managers in the cluster. However, beyond this limit, there is no implicit guarantee of high availability. Moreover, the biggest risk is the failure of the Docker machine on which the Swarm manager node is running because that is beyond the control of Docker Swarm. This risk increases in severity if these Docker machines are maintained on local servers in a hybrid cloud infrastructure, and there is no risk management facility offered by cloud providers for on-premises failures. Therefore, this paper presents a fuzzy intelligent system to address this issue and improve the high availability of Docker machines, particularly, if they are running the Docker Swarm managers locally. This fuzzy intelligent system observes the health of Docker machines, indicates their current status and generates alerts for underuse or overuse states. Therefore, an appropriate action can be taken to avoid any failures. The paper proposes and illustrates the possible solutions which can be applied depending on the underuse or overuse state of the machine. This is a generic solution and in addition to the Docker-based hybrid cloud infrastructure, it can be used with any other hybrid cloud infrastructure for improving the high availability of virtual machines.