Modified Ant Colony Placement Algorithm for Containers

Abstract

Container is an evolving lightweight virtualization innovation that attempts to perfectly capture a function and its library dependencies to be executed seamlessly at the operating system level without pre-installations or s/w setup. Placement of containers at the appropriate platform is essential in the utilization optimization of resources in cloud infrastructures. Efficient resource utilization can be achieved only when the containers are optimally mapped to VMs. Poor placement may cause a bottleneck in the cloud if VMs are loaded heavily and this may affect the response time of a given set of tasks. The ant colony optimization technique was used to schedule tasks and containers on VMs and PMs in the cloud. The disadvantage of typical ACO is its tendency to schedule tasks to the most used (high pheromone intensity) node. If the node is carrying a big load it will have an issue of overhead. By tracking preceding scheduling, this hassle could be solved by lowering the processing time and tracking load on each VM. With concerning the challenges and difficulty of the container placement, this paper proposes Modified Ant Colony Optimization Technique (MACO) for the placement of containers. The new proposal takes into consideration the scheduling history to enhance the scheduling decision. The results of MACO are compared with the basic Ant Colony Optimization technique (ACO) and First Come First Serve algorithm (FCFS). The experimental results show that the MACO is better than FCFS and the basic ACO in terms of response time and throughput.