An Optimal Cost-Efficient Resource Provisioning for Multi-servers Cloud Computing

Recently, the emergence of the cloud computing offers innovative services models to share computing resources through the internet. However, with the ever increasing in operating costs, resources management becomes the major concerns for cloud providers. In this paper, two scheduling policies are presented to stay away from the risk of over-provisioning and under-provisioning in both quantity of servers and service rate. A cloud server farm is model as an M/M/R queuing model, such that system congestion cost and balking event loss can be estimated analytically. A cost function is developed by taking system operating cost, working mode cost, system congestion cost, etc. into consideration. Two heuristic algorithms are presented to address the problem of finding optimal solutions. Simulation results show that the optimal quantity of servers and service rate can be obtained to minimum cost. As compared to system without adopting our policies, cost reduction and waiting time improvement can be verified under the same SLA constraint in our system.