Autonomic Management Techniques For Services In Multiple Clouds Contexts

With the adoptions and applications of cloud techniques,a plenty of cloud vendors are emerging on the Internet,and the services provided by them are defined as multiple cloud services.The dynamic,various,autonomic and decentralized features intensify the complexity of multiple cloud services management and impact on satisfaction degree of users' requirements and development method of cloud applications.Therefore,in multiple cloud service area it is an open issue that how to manage multiple cloud services in an efficient way to provide service for satisfying users' requirements and to support application execution and evolution.To tackle these issues,this thesis employs MAS method to abstract and model management of multiple cloud service;investigates the dynamic and evolving features of multiple cloud services via Bohr atom model;realizes service optimal provision for users with competitive game theory;realizes continuous execution and online evolution of multiple cloud applications based on autonomous management idea of multiple cloud services and service adapter concept.The contributions of this thesis are summarized as follows:1.To overcome the management problem raised by autonomic behaviors,decentralized management and dynamic running of multiple cloud services,this thesis proposes a MAS based approach to manage services and a self-organizing technique of service autonomous aggregation and provision.Via the thought of autonomic computing and the capability of agents' context awareness and autonomous decision,this thesis proposes one autonomous management model based MAS about multiple cloud services;through self-organizing interaction and cooperation among agents,this thesis proposes the techniques and algorithms supporting aggregation,organization and provision of services and adapting to the variations of users' requirements and services.Based on public dataset,this thesis conducts a series of experiments,whose results demonstrate that: mentioned models and algorithms can autonomously manage multiple cloud services,promote the efficiency of service management,and satisfy dynamic and various user requirements.2.To solve service management and inefficient service discovery problems caused by dynamic and evolving features of multiple cloud services,this thesis proposes one service field model based on atom model and its discovery method.This thesis uses energy level concept and electron jumping mechanism to describe cloud services status and changes for service classification and discovery.Based on cloud service field model,this thesis proposes a concept of user acceptable service field to confirm the boundary and method searching services that satisfy user requirements.Moreover,this thesis proposes several algorithms for the generation and updating of service field model and user acceptable service regions.Based on QWS dataset,this thesis conducts a series of experiments,whose results demonstrate that: mentioned models and algorithms can effectively depict and manage the dynamic and evolving features,and can improve the accuracy and efficiency of service discovery.3.To reduce the dynamic and uncertain features' influence on users choice behaviors,service market share and competitive strategies making during service provision,this thesis proposes one multiple cloud service provision method based on game theory.Inspired by market economic and game theories,this thesis brings quality change into user choice behavior,proposes one competitive game model of multiple cloud service provision.In that model,vendors play game in one game space of multiple cloud service marketplace;users dynamically choose services based on service prices and quality;vendors adjust their competitive strategies based on users choice.Via GNEP and its solving algorithms,this thesis proves the existence of game solution and describes the solving process of vendors' optimal provision strategies.This thesis conducts a series of experiments,whose results demonstrate that: mentioned model and algorithms can satisfy optimal utility requirements of users and vendors.4.To satisfy continuous execution and online evolution requirements of multiple cloud applications for the variation of services and users demands,based on service Adapter and Rule,this thesis proposes one Execution and Evolution pattern of multiple cloud application—ARE2 and develops one Autonomous management platform of multiple cloud Services—AutoService.Via adapter concept and MAS management model of multiple cloud services,this thesis proposes one ARE2 pattern,which shields the heterogeneous of services and quickly switches services of multiple clouds application on runtime.Moreover,based on DevOps,this thesis also proposes one application development process for its online evolution under the changes of services and user requirements.AutoService verifies the effectiveness and feasibility of mentioned methods and techniques.The simulated case illustrates that ARE2 pattern and AutoService platform can satisfy runtime execution and online evolution requirements of multiple cloud application.