Performance Modeling And Analysis Of SaaS Cloud Service Network Based On Complex Network

Cloud computing services provide service oriented architecture, open standards, scalable system, so the cloud services can provide reliable and on-demand services for users, multi-tenancy is the key technology of cloud computing. The implement of multi-tenancy in the application layer can support higher tenants density and lower cost of software management. The implementation method for SaaS cloud services is through multiple tenants share application instances.This way can easily causes interference in performance between multiple tenants. Instances composed complex network through the business association. Dynamic deployed technology makes the network of application instances can dynamically changes. Different topology structures have different performance. Because of the application instances network is indigestible, nondeterministic, it has brought many difficulties for the application instances deployment and resources scheduling of the cloud services.The complex network theory provides a new perspective for the study of large complex systems. By analyzing the characteristics of cloud services network, this thesis construct a dynamic execution network through regard application instances as a node and regard the interaction between the nodes as edges, modeling the dynamic execution process of the task. Several related performance indicators are defined. On this basis, the performance of the application instances network is analyzed. First, this thesis makes an analysis of the different performance of the different topology structure in complex network theory. Mainly includes random network, regular network, small-world network and scale-free network. Secondly, analyzes the influence of dynamic deployment and services restructuring of cloud services, including the deployment of new nodes and the reconnection of edges. Then the resources distribution methods of node are analyzed, which includes Allocate resources based on betweenness of the node and resource dynamic deployment. Under conditions of allocating resource based on degree, the performance bottleneck of scale-free network is analyzed by comparing the different resources deployment strategy.Cloud service network can dynamic deployment multiple instances in run-time. If an application has a large demand of business needs, task would have different transfer method. Three tasks transfer strategies which includes based on degree, betweenness, and closeness are analyzed in this thesis. Finally, the performance of the entry node in different degree distribution is analyzed. Relevant conclusions have an important significance for the dynamic deployment and resources scheduling of cloud services.