Research On Model And Key Realization Issues Of Cloud Computing Federation

With the widely adoption and further study of cloud computing technologies, there is an imperative demand to incorporate the resources from different cloud computing platforms to provide service for the users. Cloud Computing Federation is a computing mechanism which can federate the resources from different cloud computing platforms to provide a unified computing image for all the users. It has become one of the hot topics in Cloud Computing study.Cloud Computing Federation has the typical characteristics of the Complex System, in this dissertation, based on Complex System theory's viewpoint and methodology; in order to provide optimized and high quality computing resources for the user, we studied the architecture model, resource management, realization and load balancing problems of Cloud Computing Federation.Firstly, the architecture of the Cloud Computing Federation has significant effect on the design, realization, and the efficiency of resource management in Cloud Computing Federation. Currently, only a few works are done on this issue, some of the works only proposed preliminary models, and the others have not deliberately considering the actual resource management requirements in the Cloud Computing Federation. As the components and the running of the Cloud Computing Federation has the typical characteristic of the Complex System, therefore, based on the Emergence theory and Complex Network theory, we use the Complex Network theory to model the architecture of Cloud Computing Federation, and presented in this dissertation a Cloud Computing Federation architecture model called Region and Complex Network Theory Based Architecture for Cloud Computing Federation (abbreviated as RCNBA) with the detailed design of the resource announcement and resource discovery mechanisms. Compare with the existing resource management mechanism in P2P and Grid, a Task Manager in RCNBA can use the historical information and current status get from other Task Managers for the evolution of the complex network which is composed of Task Managers, thus meet the requirements of the resource management well in the complex Cloud Computing Federation, and it also has more advantages in the optimization of the resources, resource discovery speed, fault tolerance and adaptive ability. The result of the model experiment confirmed the advantage of RCNBA in resource discovery performance.Secondly, because there is not a standard available for Cloud Computing and the Cloud Computing platforms are heterogeneous, portability and interoperability is impossible between applications which running on different Cloud Computing Platforms, thus handicaps the realization of Cloud Computing Federation. In the existing works on Open Cloud Computing Federation, some of them only propose for the use of the existing technologies in Grid, and the others only realize the federation of enterprise clouds of the same kind, no work is found on the realization of the federation which composed of heterogeneous clouds. Based on the architecture RCNBA, we propose the Mobile Agent Based Open Cloud Computing Federation (abbreviated as MBOCCF) mechanism in this dissertation, it combines the advantages of Mobile Agent and Cloud Computing, and intends to realize portability and interoperability in the existing heterogeneous Cloud Computing platforms in current situation where no standard is available, it can be a beginning of Open Cloud Computing Federation and become a future part in Cloud Computing. We also analyzed the practicability and the performance of MBOCC, one prototype on single computer and another one on many virtue machines are designed for the test of the performance and the feasibility of MBOCC, respectively. The result of the experiments manifested MBOCCF's advantage and its feasibility.Once more, load balancing has great effect on the service quality and the efficiency of the Cloud Computing Federation, Existing Cloud Computing platforms provide simple load balancing mechanisms in the same cloud and most of them are centralized; and the existing mechanisms for load balancing in P2P and Grid can't meet the requirements of the load balancing in the Cloud Computing Federation. Currently no work is found on the load balancing in Cloud Computing Federation. Therefore, in order to realize the load balancing provided by the Cloud Computing Federation itself, we propose in this dissertation an Ant Colony and Complex Network Theory Based Load Balancing Mechanism for Cloud Computing Federation (abbreviated as ACCLB) mechanism which is guided by the Complex System theory, this mechanism realizes a distributed load mechanism based on the Emergence theory, it improves many aspects of the related Ant Colony algorithms which proposed to realize load balancing, furthermore, this mechanism takes the small-world and scale-free characteristic of Complex Network into consideration, thus a better performance on load balancing is gained. After the qualitative analysis, we design a prototype to compare the performance of ACCLB with related mechanisms under the situation where the initial network is random network, small-world network and scale-free network, respectively, and the experiments results confirmed the performance advantage of ACCLB, moreover, the simulation also manifested that if the processing capacity of the node is a random value, then the three kind of networks will all evolve into scale-free networks, the value of the degree distribution exponent y is between 2.31-2.63.Finally, in some applications, the Cloud Computing Service providers don't understand the realization specifications and load balancing strategies in the applications of the users, moreover, the status of the resources in the cloud computing platform is dynamically changing, the ants which represent the applications should adaptively change its deed according to its strategy and the change in the environments, so that to discovery the resources quickly, and deploy the application in the found resources and execute load balancing on them. Therefore, an application sponsored load balancing mechanism is needed. For this purpose, we propose A Customizable and Adaptive Load Balancing Mechanism Based on Ant Colony and Complex Network Theory in Open Cloud Computing Federation(abbreviated as CAACCLB) in this dissertation, and a atomic and totally ordered group communication mechanism which is compulsory for the ants is also given, the performance analysis manifested the advantages of this two mechanisms.Based on Complex System theory's viewpoint and methodology, to the problems that lack of studies on the architecture of the Cloud Computing Federation and the difficulties in the realization of the Cloud Computing Federation under the current technologies, we propose a model called Region and Complex Network Theory Based Architecture for Cloud Computing Federation, and a Mobile Agent Based Open Cloud Computing Federation mechanism. Based on these work, two load balancing mechanisms are presented for two load balancing modes, respectively. Our work is a valuable reference for the realization and study of Cloud Computing Federation. Furthermore, the work in this dissertation can also be used as a useful reference as technology and methodology for the study of the load balancing mechanisms in large scale distributed computing systems.