QoS-aware dynamic virtual resource management in the cloud

Virtual resource management is a key issue in cloud computing paradigm. This paper focuses on the long term/short term virtual machine rental problem. A learning algorithm based on statistical learning techniques for resource requirement is proposed, and a dynamic virtual machine rental algorithm is given. These algorithms minimized the operational cost while preserving predetermined Quality-of-Service (QoS) specification. A simulation platform is constructed, and end users' arrivals are generated according to a Markov Modulated Poisson Process ( MMPP) to simulate the load in the real world. The effectiveness of the proposed algorithms are tested via extensive numerical studies.