Building IaaS Clouds and the art of virtual machine management

Abstract

Nowadays, Infrastructure as a Service (IaaS) clouds are considered a viable solution for the on-demand provisioning of computational resources. Since its popularization in 2006 by the Amazon Elastic Computing Cloud (EC2), the IaaS paradigm is being adopted by many organizations not only to lease resources from a Cloud provider, but also to implement on-premise IaaS Clouds. The former, and original usage, is usually referred as Public Clouds while the latter is commonly named Private Cloud. Public as well as private clouds have rapidly evolved since the advent of the IaaS cloud paradigm. The public IaaS market has been enriched with multiple providers each one with different price model and offers, Cloud interfaces and APIs and even a disparate features. The private ecosystem is not different and multiple technologies both open-source and private can be used today to build on-premise clouds. Again the features, characteristics and adoption levels greatly vary among these technologies. In this talk we will review the main characteristic of the IaaS model for resource provisioning. In particular, we will analyze one of the key components to fulfill this view, the Virtual Resource management system or Cloud OS. The Cloud OS manages the physical and virtual infrastructures, and commands and controls the service provisioning, by orchestrating the deployment of virtual resources. This analysis is performed using OpenNebula as a paradigmatic example. In particular, the presentation will analyze the challenges of managing compute, storage and networking resources in a distributed environment; present the VM management model adopted by OpenNebula; and study the operation of OpenNebula in clouds environments. Finally, the talk will include the description of the Hybrid Cloud computing model, a paradigm that combines on-premise private clouds with the resources of public clouds. This new model is not yet fully developed, and there are still work to be done before true multi-cloud installations become mature enough to be used in production environments. We will discuss some of them and present some use cases.