Adaptiveness in Structured Parallel Computations

Gabriele Mencagli received a PhD in Computer Science at the Computer Science Department, University of Pisa. His research interests are in the area of High-Performance Computing. He focuses on performance models of parallel computations and their exploitation to design control-theoretic adaptation strategies in dynamic execution environments.

Adaptiveness in distributed parallel applications is an essential feature whose importance has been assessed in research fields such as scientific computing, large-scale real-time simulations and emergency management systems. This feature is of special interest to properly and promptly respond to time-varying QoS requirements, to react to highly variable workload peaks and to deal with irregular parallel problems. Starting out from the performance predictability of structured parallelism schemes, this book describes a novel formalization of adaptive parallel modules performing structured parallel computations. The module behavior is described in terms of a Hybrid System abstraction and reconfigurations are driven using the Model-based Predictive Control approach. The book provides formal solutions to control large-scale distributed applications composed of several interacting adaptive modules. Concepts of non-cooperative games and distributed cooperative optimization are presented and applied to real examples. Finally, in the book the author discusses first validations of the proposed techniques through experiments performed in a simulation environment.