Research On Key Technologies For Cloud Service Reliability Enhancement

Recently, cloud computing has emerged as a new paradigm for offering computing as services via the Internet. Because of the flexible feature, the physical resource sharing, and the pay-as-you-go model, great attentions have been pain to cloud computing by both the industry and academia researchers. An increasing number of companies are beginning to deploy services/applications in the cloud computing environment.However, the large quantity of the virtual resources, the uncertainty of the failure events and the non-sustainable obtain of the virtual machines have badly affected the cloud service reliability. Enhancing the reliability of cloud service has become a critical and challenging research problem. Hence, taking the virtual resource supplier and the cloud application supplier as the research objects, this paper researches on key technologies for cloud service reliability enhancement, and reaches the following results:1. Checkpoint/restart is a suitable virtual resource reliability enhancement mechanism when the virtual machine hosts continuous tasks. In cloud computing environment, existing approaches cannot achieve the optimal effect because of checkpoint image-sharing neglect, datacenter network architecture neglect, and checkpoint image inaccessibility caused by node crashing. This work proposes an optimal reliability-aware, distributed image-sharing approach. We propose this checkpoint approach for minimizing network and storage resource usage of datacenter. To verify the effectiveness of the proposed approach, we extend the renowned cloud simulator Cloudsim and conduct experiments on the new simulator FTCloudsim. Experimental results show that the proposed approach not only guarantees cloud service reliability, but also consumes fewer network and storage resources than other approaches.2. Replication is a commonly used virtual resource reliability enhancement technique when the virtual machine hosts discrete tasks. However, when determining the virtual machine placement strategy, existing approaches ignore the huge network resource consumption in the recovery stage. To solve this problem, this work proposes a datacenter network topology-aware virtual machine placement and tasks rescheduling approach. To verify the effectiveness of our approach, we conduct experiment on FTCloudsim. The experimental results showed that the proposed approach consumes less network resource.3. Many cloud application suppliers purchase spot virtual machine to save the resource costs. Because the virtual machine can be revoked due to price fluctuation, this purchase model offers low resource rental cost in exchange for reliability. Therefore, the virtual resource supplier offers a checkpoint interface for the cloud application suppliers to trade-off between the monetary cost and reliability. Based on historical price distribution, we propose an adaptive cost-aware checkpoint approach to minimize the rental cost but still ensure the cloud application reliability. We provide an extensive evaluation of our approach using Amazon EC2 spot price dataset. The results on FTCloudsim showed that the proposed approach can enhance the service reliability and reduce the cost of the cloud application suppliers.