Survivability Evaluation of Disaster Tolerant Cloud Computing Systems
Bruno Silva, Paulo Romero, Martins Maciel, Armin Zimmermann · 2014
A prominent type of cloud service is the Infrastructure-as-a-Service (IaaS), which delivers, on-demand, computing resources in the form of virtual machines (VMs) satisfying user needs. In such systems, penalties may be applied if the defined quality level of service level agreement (SLA) is not satisfied. Therefore, high availability is a critical requirement of these systems. A strategy to protect such systems from natural or manmade disasters corresponds to the utilization of multiple data centers located into different geographical locations to provide the service. Considering such systems, redundancy mechanisms can be adopted to receive copies of VM images during data center operation. Hence, whenever a disaster makes one data center unavailable, the VMs can be re-instantiated in other operational data center. Modeling techniques, with a strong mathematical foundation, such as Stochastic Petri Nets (SPN) can be adopted to evaluate survivability in these complex infrastructures. This work presents SPN models to evaluate survivability metrics in IaaS systems deployed into geographically distributed data centers taking into account disaster occurrences. Using the proposed models, IaaS providers can evaluate the impact of VM transmission time and the VM backup period on survivability metrics. A case study is provided to illustrate the effectiveness of the proposed work.