Model-Based Sensitivity of a Disaster Tolerant Active-Active GENESIS Cloud System

Modern cloud computing systems are prone to disasters. And the true cost due to service outages is reportedly huge. Some of previous works presented the use of hierarchical models: fault tree (FT), reliability block diagram (RBD) along with state-space models: continuous time Markov chain (CTMC) or stochastic petri nets (SPN) to assess the reliability/availability of cloud systems, but with much simplification. In this paper, we attempt to propose a combinatorial monolithic model using reliability graph (RG) for a real-world cloud system called general purpose integrated cloud system (GENESIS). The system is designed in active-active high availability configuration with two geographically distributed cloud sites for the sake of disaster tolerance (DT). We then present the model-based comprehensive analysis of system reliability/availability and their sensitivity. The results pinpoint different findings in which the architecture of active-active and geographically dispersed sites with appropriate interconnections of the cloud apparently enhance the system reliability/availability and assure disaster tolerance for the cloud.