A Programming Model for High-Performance Adaptive Applications on Pervasive Mobile Grids

Pervasive Grids are emerging distributed computing plat- forms featuring high degrees of dynamicity and hetero- geneity. They are composed of fixed and mobile nodes, in- terconnected through wireless and wired networks. Adap- tivity is the key point for applications to efficiently exploit Pervasive Grids. We focus on High-Performance pervasive applications, such as emergency management. We present a novel programming model which allows programmers to express high-performance adaptive applications in terms of their parallel and distributed structures. In this program- ming model a parallel module can be programmed in mul- tiple versions, each optimized for a specific platform con- figuration (e.g. mobile nodes or central servers). The mod- ule is programmed to select the best version to handle/serve certain events, such as wireless link failure and user request for a better QoS. We experimentally show the efficacy of this approach for a test-based application on a Pervasive Grid.