A VM Memory Demand Prediction Algorithm Based on the RBF-Markov

The utilization of virtual memory is always undefined due to the lasting changes of the service requests and the running programs, so the utilization of the multiple VMs memory particularly complicated. In the existing virtual systems, many resource allocation plans of the VM memory lack an efficient and flexible dynamic adjustment mechanism, which is the main reason of the disharmony state of the virtual system. In the long-term running, the business peaks and valleys of each VM are not synchronous that causes that different machines need different amount of the memory resource. Moreover, with the uncertain external load changing, the service level of VM is also uncertain. So it is an important problem that adjusts the memory dynamically based on the current state. Aiming at the shortcomings of the traditional prediction algorithms, this paper proposes an objectives-oriented algorithm, based on RBF-Markov, to predict the future VMs' memory requirements.