Multiple-Backup Resource Allocation Model for Virtual Machines With Probabilistic Protection

For cloud providers, it is essential to improve the quality of service that depends on the failure probability of protection and the computing capacity cost such as backup resources. Existing studies addressed a protection approach to reduce the required backup capacity by sharing backup capacity among multiple primary resources. Still, the sharing effect is limited, and there is room to enhance it; more primary capacity should share more backup capacity for the enhancement. This paper proposes a model that minimizes the required backup capacity. This model ensures that the backup failure probability, or the probability of unsuccessful backup, of primary resources during multiple simultaneous physical machine (PM) failures does not exceed a given value. In this model, when a PM containing virtual machines (VMs) fails, the VMs in that PM can be recovered by available backup resources in backup PMs. By setting the priority of protecting VMs and backup PMs that each VM is protected by, we obtain the backup failure probability when all available backup resources are used for backup. We introduce heuristic approaches to allocate backup capacity and prioritize the protection to minimize the required capacity while satisfying a given backup failure probability. We introduce a priority policy and a computation policy to reduce the computation time and to be able to deal with larger-size problems. The proposed model can reduce the total required backup capacity compared to the baseline models. We can make the proposed model possible to protect primary resources in larger-size problems.