Adaptive resource management for balancing availability and performance in cloud computing

Security, availability and performance are critical to meet service level agreements in most Cloud computing services. In this paper, we build on the virtual machine technology that allows software components to be cheaply moved, replicated, and allocated on the hardware infrastructure to devise a solution that ensures users availability and performance requirements in Cloud environments. To deal with failures and vulnerabilities also due to cyber-attacks, we formulate the availability and performance attributes in the users perspective and show that the two attributes may often be competing for a given application. We then present a heuristics-based approach that restores application's requirements in the failure and recovery events. Our algorithm uses Markov chains and queuing networks to estimate the availability and performance of different deployment contexts, and generates a set of actions to re-deploy a given application. By simulation, we show that our proposed approach improves the availability and lowers the degradation of system's response time compared to traditional static schemes.