Elastic Highly Available Cloud Computing

High availability and elasticity are two the cloud computing services technical features. Elasticity is a key feature of cloud computing where provisioning of resources is closely tied to the runtime demand. High availability assures that cloud applications are resilient to failures. Existing cloud solutions focus on providing both features at the level of the virtual resource through virtual machines by managing their restart, addition, and removal as needed. These existing solutions map applications to a specific design, which is not suitable for many applications especially virtualized telecommunication applications that are required to meet carrier grade standards. Carrier grade applications typically rely on the underlying platform to manage their availability by monitoring heartbeats, executing recoveries, and attempting repairs to bring the system back to normal. Migrating such applications to the cloud can be particularly challenging, especially if the elasticity policies target the application only, without considering the underlying platform contributing to its high availability (HA). In this thesis, a Network Function Virtualization (NFV) framework is introduced; the challenges and requirements of its use in mobile networks are discussed. In particular, an architecture for NFV framework entities in the virtual environment is proposed. In order to reduce signaling traffic congestion and achieve better performance, a criterion to bundle multiple functions of virtualized evolved packet-core in a single physical device or a group of adjacent devices is proposed. The analysis shows that the proposed grouping can reduce the network control traffic by 70 percent. Moreover, a comprehensive framework for the elasticity of highly available applications that considers the elastic deployment of the platform and the HA placement of the application’s components is proposed. The approach is applied to an internet protocol multimedia subsystem (IMS) application and demonstrate how, within a matter of seconds, the IMS application can be scaled up while maintaining its HA status.