| As cloud computing becomes more important,naturally,end users expect the services offered by cloud computing to deliver at least the same service dependability as traditional models provide;meanwhile,the service providers and infrastructure suppliers also look forward to the prospect the dependable cloud services bring more benefit and a better reputation.However,due to the heterogeneity of the cloud environment,the diversity of cloud services and complexity of virtualization technology,it is more difficult to discover the real cause of the failures.Though numerous mechanisms or methods have been approached in different scenarios and served the purpose of enhancing the dependability of the systems,this research field still requires theoretical foundations,appropriate models,effective design paradigm,in-depth experimentations,and practical implementations so as to construct highly-dependable cloud systems.Henceforth,the main object of the thesis focuses on the ways to enhance the dependability of cloud service in the specific application scenario of the QoS-aware mechanism.Firstly,it proposes and formalizes the specific QoS and RoS model of cloud service,and constructs the mapping between them;furthermore,for the sake of dependability enhancement,it puts forward useful and various solutions from the following four different layers:(1)Focusing on the requirement for retrieving the dependable cloud service,we propose an extended UDDI—CloudUDDI,which still includes the general functions of the UDDI and expends resource description and measurement of cloud environment.It can assist the consumers to find high quality cloud services more accurately via QoS-aware and RoS-aware.According to the experiment,the results show that CloudUDDI can effectively support the registration and lookup of cloud services,and facilitate the users' choices of suitable replacements service tackling with the service failure conditions.(2)Robust application platforms generally have some type of failure detection,reporting and recovery mechanism.Combined with the self-healing mechanism and AOP technology,we propose a framework for enhancing dependability of cloud services,which can allocate the dependable strategies of cloud services in the form of “pluggable” in logic layer.Using fault injection method for simulation,the result shows that this framework can reduce the possibility of cloud service failure validly and enhances the dependability of the cloud service.(3)Provisioning and scheduling resource of virtual machines effectively could deal well with a variety of task requests,which is a kind of “soft” fault tolerance in essence.The paper illustrates theway of modeling load status of the virtual machine instance,estimating the resource utilization ratio of the virtual machine cluster quantitatively,designing the task scheduling algorithm and elastic scaling algorithm,targeted at the tasks distribution and the elastic scaling of the virtual machines cluster respectively.Through the simulation using CloudSim platform,the result shows that this novel method has better load balance degree and complement time,compared with other common algorithms such as round robin and green load,which ultimately achieves the purpose of optimal load balancing in premise of ensuring QoS.(4)The paper demonstrates a QoS-aware framework for fault tolerance of cloud service,which can choose suitable fault tolerance strategies via probing the status of QoS of cloud service.An error generation mechanism is designed based on Urn and Bull model in Hidden Markov modeling theory,which can simulate real causes leading to service failure,as well as verify the effectiveness of the framework.The results show that the fault tolerance framework has valid fault-tolerant capability and enhances the availability and reliability. |
PDF Full Text Request