DOI: 10.5176/978-981-08-5837-7_219
Authors: Xiaofeng Jiang, Hongsheng Xi, Qing Chen, Jun Li, Fengbin Li
Abstract:
Resource scheduling strategy plays an important role in cloud computing. How to appropriately distribute resources to multiple services when the cloud network stays in different load conditions is the main work of the scheduling strategy. By analyzing the behavior of the cloud network with multiple services, we consider the cloud as a resource pool, and each kind of the services in the cloud network is converted to an M/G/1 queue, and all the queues make up a queuing network, so we can describe the cloud as a model of an open queuing network. The best service rate of one service will be given by our scheduling strategy optimization algorithm to make the profits of the cloud maximum and guarantee the Quality of Services(QoS). In this paper, first, the measurement of QoS accord with the queuing network model is given to limit the range of the scheduling policy. Then considering the model of the open queuing network, we add another queue which represents the source of the customers in the model, and convert it into a closed queuing network. An iterative algorithm is adopted to iterate the profit function of the cloud, which consists of running costs, maintenance costs and the rewards from the customers. Meanwhile, we use the QoS torestrict the iterating results. There are some matrix operations in the iteration, and a simplified method in order to reduce the computational load caused by the matrices of high dimensions has been given. At the last of the iteration, a proof of the convergence of the iteration has been given. Finally, simulations and numeral results are provided. Adequate experiments are carried out to observe the profits of the simulated cloud network. The results from different initial data show us that the convergent strategy not only makes the profits of the cloud maximum, but also guarantees the customer's QoS.
Updating...