Behavioural Skeletons in FastFlow

This thesis work consists in the implementation of a version of the Behavioural Skeletons (BS) within the structured parallel programming framework FastFlow (FF). Therefore design, implementation and experimentation are here considered and discussed. Furthermore, with the introduction of the BS in FastFlow, we implement a fully functional Autonomic System for run-time optimization of non-functional concerns. Extensive details are given for the design and implementation choices of the autonomic components. Moreover we discuss design and implementation choices for modifications to the already present algorithmic skeletons of FF. The above mentioned variations give the skeletons dynamic features, permitting run-time changes of their structure. As for the management subsystem, we discuss the realization of sensors and actuators (Autonomic Controller) for the normal FF skeletons and the different available models for the management components (Atonomic Managers). Experiments are conducted to demonstrate the features of the newly extended FastFlow framework, with functional experiments covering the majority of the implemented components and an example of run-time optimiziation of a composed complex Behavioral Skeleton structure. In conclusion, we have demonstrated succsessful design, implementation and experimentation of Behavioural Skeletons, typical constructs of distributed computing, in the different context of parallel computing. Which leads to a fully functional autonomic system model for the FastFlow framework