Research And Design Of Fault Injection Platform For Cloud Computing System

With the development of cloud computing technology, a wide range of cloud computing platforms have emerged out, a various of research institutions and large companies have launched their own cloud platforms, to provide users with convenient and scalable cloud computing services. Cloud computing platform is the core of the cloud infrastructure, cloud services in all applications need to be deployed on cloud platforms, cloud platform can manage the underlying hardware resources such as network resources, storage resources and comp uting resources. The virtualization layer and can put a physical hardware into multiple virtual devices to the top of the software. Cloud computing includes muti-layers, Iaa S, Paa S and Saa S layer. However, with the increasing complexity and scale of the cl oud systems, cloud system failures and errors that may occur during operation, more and more serious failures will give companies and users serious losses.This paper reviews the robustness of the cloud computing system this problem, build a cloud computing system for fault injection platform with a distributed architecture, based on the typical hierarchical structure of cloud systems, the fault injection into multiple levels, respectively distributed computing layer, the virtualization layer and the cloud management stacks layer. This paper designs a unified fault injection framework, standardized the fault injection process, giving testers a variety of interfaces to facilitate testing conducted a species is a graphical interface to the real-time injection, one is scripted batch injection. And simple fault can be detected by our system and we can collect informations on fault injection.After that, this paper selects typical practical systems from three levels of fault models, distributed computing layer, the virtualization layer and the cloud management platform stack layers, then we investigated their features and fault tolerance mechanisms, and designs tools for the injection fault model and the specific characteristics of this fault.Finally, we use Cloud Stack, Xen Server, KVM, Hadoop, Spark composition to build a cloud system fault injection testing laboratory. First of all, we conducted fault injection experiment for each levels, and then the different levels of cross-level fault injection experiments. Through experiments, we not only verify some system of automatic fault tolerance, but also on some of the deficiencies of the system and put forward suggestions for improvement.