Multi-Objective Scheduling using Logistic Regression for OpenStack-based Cloud

Abstract

Abstract Efficient resource management in Cloud data center is a challenging problem. Resource monitoring has been reported as a tedious task in multi-tenant data centers. Implementing an effective technique for monitoring and scheduling of the cloud resources is the need of today’s data centers. OpenStack is an open source distributed cloud computing platform. Default scheduler of OpenStack implements worst fit algorithm which assigns VM to host with only maximum available RAM which is one-dimensional so it leads to inefficient usage of resources. This paper proposes a new model considering multiple parameters such as memory, CPU utilization and current workload of the hosts. This work proposes for the use of machine learning classifier to classify whether host is overloaded or under loaded. Proposed scheduler uses the decision provided by the classifier and implements best fit algorithm which schedules VM to host by considering multiple parameters. Experimental results show that proposed algorithm outperforms default scheduler and it leads to efficient usage of underlying resources of hosts.