Reliability Analysis of Cloud Computing Systems with Different Scheduling Strategies under Dynamic Demands

Abstract

The cloud computing system (CCS) is with large scale and complex structure. Benefited from its enabling technology, virtualization, the CCS has the characteristics of elastic, providing on-demand service according to resource scheduling strategy. Different resource scheduling strategies lead to different operation hours, which have different effects on the reliability of cloud computing infrastructure. In this paper, the reliability of CCS is calculated based on the reliability of its physical servers and virtual machines (VMs), and the system reliability with two mainstream resource scheduling strategies is compared. In the process of analysis, the time is divided into several small time periods, and the system demand in each time period is assumed to be the same. The demand is allocated to VMs considering their states according to the resource scheduling strategy. The four states (i.e., idle, operation, failure and migration failed) of VMs and the state transitions are analyzed. A reliability simulation method based on Monte Carlo is provided, and it considers the characteristics of elastic scheduling, common-cause failure and fault-tolerant migration behavior of the CCS. The effectiveness of our method is verified by our illustrated case, and the system reliability with the load balancing and the energy based scheduling strategies are compared. The results show that the reliability of the CCS is higher with the energy based scheduling strategy than the one with the load balancing based scheduling strategy.