Service-oriented Resource Orchestration – A Resource Allocation Approach

Abstract

This thesis is primarily concerned with three main topics in cloud platforms using OpenStack as a case study: allocation of resources to meet the demands of a service requested by remote end-user, migration of virtual machines (VMs) instances to offload the overloaded compute nodes, and monitoring of utilized resources. The overall framework architecture consists of three subsystems: 1) An orchestrator that enables to automate resource management and provisioning in OpenStack, 2) sFlow-based subsystem to monitor resource performance counters of OpenStack compute nodes, and 3) A resource utilization-based subsystem for dynamic VM migration in OpenStack. The proposed orchestrator manages and provisions resources by: 1) exploiting application program interfaces (APIs) provided by the cloud provider in order to control/manage/allocate storage and compute resources; 2) interacting with software-defined networking (SDN) controller to get the details of the available resources, and instructing the changes/rules to manage the network based on the cloud service requirements. For resource allocation, an algorithm is proposed, which allocates resources on the basis of unutilized resources in a pool of virtual machines. An algorithm has been taken into account for comparison with the proposed resource allocation algorithm. The experiment results show that the considered algorithm is outperformed by the proposed algorithm. Furthermore, a monitoring system has been implemented, which collects and stores samples of the performance counters related to infrastructural resources, such as CPU performance, memory faults, and network performance. A framework is proposed for dynamic VM migration in a cloud computing platform. The framework implements the proposed overload detection, VM selection, and VM allocation algorithms for dynamic VM migration in clouds. With the help of experiments, it is shown that the proposed algorithms outperform the algorithms that are considered for the purpose of evaluation.