Persistence, discovery, and generation of viable cloud application topologies

Abstract

Cloud computing is gaining popularity among application developers with each passing day because of the many benefits introduced by the cloud paradigm. Also the number and range of available cloud services is continuously increasing and hence the application developers are willing to migrate complete or partial applications to cloud environment. Applications can be designed to be run on the cloud, and utilize its technologies, or can be partially or totally migrated to the cloud. The application’s architecture contains three layers: presentation, business logic, and data layer. Each of this layers can be deployed on a different cloud service basing on the requirements and the constraints and it is also possible to deploy a layer of the application on an on premise physical server if required and topologies need to be designed for the deployment of the application. There are standards like TOSCA, or approaches like GENTL and MOCCA which allow for the modeling and management of application topologies. Cloud application topologies can be defined as typed labeled graphs constituted by a set of nodes, edges, and labels. Nodes represent the application components, while the edges depict the relationship among them and sub-topologies can be built by grouping multiple nodes. A cloud application viable topology is a feasible distribution of the application components according to the depicted or discovered application independent sub-topologies. Some objectives like cost, performance, etc also need to be considered in order to help analyze the fitness of the services for the desired operation. This Master Thesis focuses on optimally distributing application components across cloud offerings efficiently. More specifically this work deals with providing a topology modeling framework capable of supporting the following three fundamental aspects: (i) leveraging existing technologies and mechanisms for analyzing the different aspects of the evolution of cloud application topologies, (ii) design and develop the concepts and mechanisms towards dynamically discovering and constructing cloud application viable distributions (viable topologies) specifications (typically XML representations) in an optimal manner, and (iii) developing the visualization means within an existing topology modeling environment.