A Novel Fault-Tolerant Container Scheduling Algorithm in Docker Based Cloud

Abstract

Docker Container is newly introduced technology in cloud Computing since last few years. Docker Swarm is tool for scheduling and managing individual Container for microservice application within a cluster of private cloud and public cloud. Docker swarm has three scheduling strategies spread, binpack and random. Currently, only Spread scheduling strategy is supported in docker swarm. Spread strategy is based on single objective to spread containers from each service over as many nodes as available, as evenly. But when there's a choice between suitable nodes for the next containers, preference is given to the node with the fewest total containers. However, when any node is failed it will place its running containers evenly on available nodes and in this situation some nodes are overloaded and some node will become underloaded. Other drawback of spread strategy is while creating container with memory reservation if suitable nodes are not available it will place all containers on same nodes. In this paper we proposed new scheduling strategy that will solve the problem and no node will be overloaded and underloaded after failure of nodes in cluster.