Utility-based analysis of evolving cloud application topologies

Abstract

Today, Cloud consumers have access to a wide spectrum of Cloud offerings. On the one hand, this is a profitable situation, since there is a larger spectrum of possibilities to migrate the application to the Cloud. However, on the other hand, consumers face the challenge of selecting the offering that promises the highest benefit. The challenge even grows larger when taking into consideration the possibility to distribute the application components. Approaches like TOSCA support developers in in the portable description of composite Cloud applications and tools provide the selection of the most cost-effective Cloud offerings that fulfill a set of requirements. Besides that, there is a lack of decision support that goes beyond the mere look on Cloud offerings' technical data and operational costs. Developers should be supported with the necessary mechanisms and tools towards evaluating and analysing the trade-off between different aspects and involve different stakeholders' interests. Moreover not only isolated Cloud offerings should be evaluated, but also the outcome of applications under a distributed deployment with respect to evolving workloads. Since utility functions facilitate the analysis of users' satisfaction, i.e. performing the trade-off between different aspects, this thesis presents a concept that uses utility functions in order to evaluate applications' topologies. This concept makes application distribution alternatives comparable. Based on this concept, a utility calculation framework is specified. The framework provides support for creating customized utility functions, calculates the utility of alternative topologies, and offers decision support. Furthermore, this thesis presents a prototypical implementation of the utility calculation framework, which is further evaluated using a realistic application and data.