Section: New Results

Cloud Computing

Participants : Elena Giachino, Michael Lienhardt, Tudor Alexandru Lascu, Jacopo Mauro, Gianluigi Zavattaro.

Languages for cloud applications

To foster the industrial adoption of virtualized services, it is necessary to address two important problems: (1) the efficient analysis, dynamic composition of services with qualitative and quantitative service levels and (2) the dynamic control of resources such as storage and processing capacities according to the internal policies of the services. Current technologies for cloud computing, addresses these problems at deployment and run time. The ENVISAGE project and the position paper [20] proposes, on the contrary, to overcome these problems by leveraging service-level agreements into software models and resource management into early phases of service design.

Models for cloud application deployment

Cloud computing offers the possibility to build sophisticated software systems on virtualized infrastructures at a fraction of the cost necessary just few years ago, but deploying/maintaining/reconfiguring such software systems is a serious challenge. The AEOLUS project, aims to tackle the scientific problems that need to be solved in order to ease the problem of efficient and cost-effective deployment and administration of the complex distributed architectures which are at the heart of cloud applications [25] . In particular, it is necessary to define appropriate models for the representation of the interdependencies among the software components of a cloud application as well as declarative languages for the specification of the desired application configuration. We have proposed [31] a model for the representation of the component lifecycle and of its dependencies/conflicts with the other components. Based on such model, we have defined a sound and complete algorithm that efficiently computes a deployment plan (i.e. a sequence of low-level component deployment actions) capable of reaching a final configuration including at least some predefined basic components [48] and we have realized a prototypical implementation of such algorithm which was proved to be effective on case-studies of realistic size (i.e. hundreds of components) [41] .