Merging of TOSCA Cloud Topology Templates

Abstract

The paradigm of Cloud Computing requires standardization to avoid vendor lock-in for its users. The Topology and Orchestration Specification for Cloud Applications (TOSCA) provides a standardization approach enabling portability of cloud services between different Cloud Computing providers. The main goal of the TOSCA specification is to enable a cloud provider and environment independent description of these services concerning structure and management aspects during their life cycle. TOSCA specifies so-called Service Templates whose structure is described by a Topology Template. The TOSCA specification provides means to model enterprise applications in a standardized way. In view of mergers and acquisitions, data center consolidation, for breaking up silo structures in IT departments and for support of modeling experts it is necessary to find concepts for analyzing already modeled TOSCA Topology Templates for similar elements and for unifying these elements and, thus, two Topology Templates. This master's thesis develops a matching concept for finding similar elements inside and between two Topology Templates by systematically exploring all different constellations TOSCA elements can take. Similar elements are indicated by the notion of a Correspondence. The matching concept is automated by developing algorithms that determine Correspondences and incorporate domain-specific knowledge via type-specific plugins. The plugins handle the matching of properties that cannot be conducted generically. Furthermore, a merging concept and appropriate algorithms are developed that utilize the identified Correspondences for unifying similar elements. All algorithms are designed with the goal of practical computational complexity. A further part of this work is the design and prototypical implementation of the extendable TOSCAMerge framework that allows for a convenient integration of type-specific plugins. The framework facilitates the assessment and manipulation of the determined Correspondences by domain experts prior to merging. A set of example TOSCA Service Templates for testing the different matching and m erging cases complements the implementation.An extensive evaluation of the concepts and algorithms reveals the algorithms' greedy properties including local optimality but quadratic computational complexity in most cases.