Section: New Results

Cloud Computing

Participants : Roberto Amadini, Maurizio Gabbrielli, Elena Giachino, Saverio Giallorenzo, Claudio Guidi, Cosimo Laneve, Michael Lienhardt, Tudor Alexandru Lascu, Jacopo Mauro, Gianluigi Zavattaro.

Cloud application deployment

Configuration and management of applications in the cloud is a complex task that requires novel methodologies and tools. In [16] we have performed a foundational study of the complexity boundaries for the automatic deployment problem, showing that in the general case this problem is undecidable, it is decidable but non-primitive recursive if capacity constraints are not taken into account, while it turns out to be polynomial time if also conflicts between software components are not considered. Starting from these foundational observations, we have investigated the exploitability in this specific context of state-of-the-art constraint optimization techniques, a well established approach for the modeling and solution of complex optimization problems. In particular, in [23] , [24] we have studied how the "portfolio technique" approach can be applied to optimization problems, combining and exploiting the performances of existing solvers to get a global, more robust and fast solver. Encouraged by these results, we have developed SUNNY-CP [13] , [22] : a porftolio constraint solver for constraint satisfaction and optimization problems. SUNNY-CP has proven to have remarkable performances, ranking 4th in the annual MiniZinc challenge (i.e., the international competition to evaluate the performances of constraint solvers) and receiving a `honorable' mention by the challenge organizers.

Cloud resource management

The management of cloud resources from client programs requires the definition of Application Programming Interfaces (APIs) that expose specific functionalities to external invokers. Programs can be built that compose existing APIs in order to obtain new functionalities. However API composition easily becomes a frustrating and time-costly task that hinders API reuse. The issue derives from technology-dependent features of API composition such as the need of extensive documentation, protocol integration, security issues, etc.. In [39] we introduce the perspective of the API-as-a-Service (APIaaS) layer as tool to ease the development and deployment of applications based on API composition, abstracting communication protocols and message formats. We elicit the desirable features of such a layer and provide a proof-of-concept prototype implemented using a service-oriented language.

Another critical aspect in this context deals with the problem of dynamic reallocation of resources. In [38] we study a type-based technique for modeling and analysis of systems in which concurrent object-oriented programs dynamically create and move resources. The type of a program is behavioural, namely it expresses the resource deployments over periods of (logical) time. Our technique admits the inference of types and may underlie the optimisation of the costs and consumption of resources.