Comparison and Evaluation of Open-source Cloud Management Software

Abstract: The number of cloud management software related to a private infrastructure-as-a-service cloud is increasing day-by-day. The features of the cloud management software vary significantly and this creates a difficulty for the cloud consumers to choose the software based on their business requirements. An example of the problem is choosing software with a power management feature. The power management feature is used to increase the efficiency of energy consumption by consolidating virtual machines together and turning off unused physical servers, which is not provided by many cloud management software. OpenNebula is one of the most widely used open-source cloud management software among research institutions and enterprises. However, the performance characteristic of OpenNebula is not well studied in the existing literature. An example of the problem is choosing a hardware configuration to run OpenNebula for the research institutions and enterprises. The first objective of this thesis is to develop a framework for comparing features of various cloud management software. For developing this framework, existing works are reviewed. The cloud management software is installed on the KTH LCN testbed for hands-on experience. Both the open-source and the commercial software are analyzed for developing the framework. The major contribution related to the framework is identifying features provided for the commercial software that are not available for the open-source software. The features are: (1) co-location of VMs is running a group of VMs on the same physical machine (for example, if the web server VM has to access the application server VM for getting the web pages, they can be placed on the same physical machine); (2) anti-co-location of VMs is not allowing a pair of VMs to run on a single physical machine (for example, the primary and back-up web server VMs should always run on the different physical machines); (3) the resources of the physical machines can be combined (e.g., number of CPU cores, physical memory) as a resource pool and compartmentalized into an organizational structure (e.g., HR, development, testing, etc). The second objective of this thesis is to evaluate the performance of the OpenNebula cloud management software. For the performance evaluation, existing works are reviewed to identify the metrics, and the OpenNebula cloud management software is installed on the KTH LCN testbed. The performance of the OpenNebula software was evaluated for different virtual machine operations, virtual machine types, number of virtual machines and change in load of the system. The major lessons learned related to the performance evaluation are: (1) the duration for the live migration does not change with the load; (2) the duration for the live migration increases linearly as the memory assigned to the VM increases; (3) the duration of the add and delete operations increases linearly as the number of VMs increases.