Section: New Results

Design and Runtime Support for Cloud Computing

In 2014, we obtained some new results in the domain both of the design and the runtime support of distributed applications for multi-cloud systems. The purpose is to deal with applications that span across several different cloud systems. Several reasons justify such a goal. For example, in order to avoid the so-called vendor lock-in syndrome, cloud application stakeholders need to be able to migrate as easily as possible their assets from one cloud system to another one. Other examples include the possibility of introducing diversify and fault-tolerance by deploying applications on different cloud systems, or hot migrating applications where computing resources are less expensive.

For the design of multi-cloud systems, we proposed a solution based on software product lines (SPL)  [90] and ontologies. In order to specify the variability of such environments, we extended SPL with attributes, cardinalities, and constraints. In order to enable the evolution of these environments, we provided an automated support for maintaining the consistency based on constraint programming. Finally, we proposed an ontology based approach to bridge the gap between the concepts and artefacts defined by different cloud systems. This global solution is the result of the PhD thesis of Clément Quinton [16] that was defended in October 2014, and has been partially supported by the FP7 PaaSage project (see Section  8.3 ).

For the runtime support of multi cloud systems, we proposed the soCloud platform. This solution enables to deploy, execute and manage an application that spans on several different cloud systems. soCloud tackles the challenges of portability, provisioning, elasticity, and high availability. soCloud defines a component-based and service-oriented architecture that provides an unified view of a set of cloud systems. soCloud is the result of the PhD thesis of Fawaz Paraiso [15] that was defended in June 2014. soCloud is implemented on top of the FraSCAti platform (see Section  5.2 ).