Optimizing Scheduling in Container Based Cloud Architecture

Abstract

quotAt present, cloud computing is the most essential model in computing technology. Cloud computing has received a great deal of attention because of its focus on providing fast and reasonable computing resources for customers. Distribution and management of limited cloud resources among users is a very challenging task. Docker-based clouds are a new emerging model based on container technology. newline newlineProject aims to address the problem of scheduling containers in cloud computing using Docker. Docker supports the spread scheduling strategy by default. Its main objective is to place containers on all available machines in the cluster. However, it does not consider load on machines. Also, if a node in the cluster fails, the spread strategy will distribute its containers uniformly across the remaining nodes. A consequence of this is that some machines in the cluster become overloaded and some machines become under loaded. newline newlineAs a result, we optimize the performance of docker clusters by taking into account the CPU usage of every machine while placing containers on machines. That will minimize the number of machines, ensure load balancing and maximize resource utilization. By using this scheduler, user can select a node with a minimum CPU utilization to place the initial docker container to improve load balancing and minimize the number of machines in the docker cluster as much as possible. In addition, we examine the failures of the servers on which the containers run. Hence, we proposed another fault tolerant container scheduling algorithm that works even when any of the nodes in the cluster fail. Both algorithms are evaluated and compared with existing spread algorithms in a real cloud environment. Compared to a spread strategy, both scheduling algorithms use minimal number of machines in the docker cluster as well as proper resource utilization. newlinequot newline newline