Design support for performance- and cost-efficient (re)distribution of cloud applications

Abstract

In the last decade the IT landscape has experienced a revolutionary change in both the research and industry domains due to the strong emergence and adoption of cloud providers and services. The increase of available cloud services together with the materialization of DevOps methodologies have forced a change in how software is engineered and operated. Nowadays, organizations can leverage a wide spectrum of cloud services to build, migrate, distribute, and even replace their application components in the cloud. In summary, the transformation of capital into operational expenditures when using private or public clouds has become a reality. The emergence of cloud services, however, brought with it a substantial complexity when designing and running cloud applications. More specifically, cloud environments are intrinsically multi-tenant, where different tenants (or cloud consumers) consume computational, network, and storage resources on an on-demand basis. This type of virtualized environment causes performance to be variable and heterogeneous, therefore having consequences on application costs. When migrating applications to the cloud, this heterogeneity causes architectural decision making to be more complex, mainly due to the limited amount of knowledge about cloud services. This work tackles the aforementioned decision making challenge, by defining and realizing a design decision support framework called SCARF to assist IT organizations to efficiently migrate their applications to the cloud, with a focus on optimizing their cost and performance. In summary, SCARF is a design and decision support framework for \textit{systematically (re)distributing applications in the cloud}. SCARF equips application architects with a design process and tools (i) to model their application architectures, (ii) to automatically discover potential cloud services to (re-)deploy their application components, (iii) to evaluate them with respect to cost and performance, and (iv) to monitor and analyze application costs and performance evolution during runtime. SCARF builds upon existing concepts and techniques in the domain of cloud application topologies, case base reasoning, similarity analysis, and utility theory, in order to simplify the steps for re-engineering and migrating traditional applications to the cloud. The technological support for SCARF, SCARF-T, comprises an integrated tool chain built upon prior standards and reusing existing technologies and tools. The evaluation of SCARF and SCARF-T show a more efficient and simplified decision making process for migrating applications to the cloud, as also shown by the publications supporting this thesis. Overall, SCARF can be considered as a starting point for enabling IT organizations with architectural decision support for migrating their applications to the cloud, with a focus on optimizing the complete spectrum of application non-functional aspects.