Utilizing Reconfigurable Hardware to Optimize Workflows in Networked Nodes

This work investigates the use of reconfigurable devices as computing platform for self-organizing embedded systems. Those usually consist of a set of distributed, autonomous nodes interacting with each other in order to solve a given problem. Several aspects of hardware-software co-design as well as partial reconfiguration are presented in order to enforce adaptivity of a node. One targeted application field for this kind of system are sensor networks in which reconfigurable devices, in this case FPGAs, can be used as computation nodes to provide services that require more computation power. To manage the available hardware resources as a whole we suggest a market-economy-like system of supply and demand. Requests, built up out of several tasks, can be posed to the collective. The goal is to gain a system able to perform simple tasks as well as very complex computations, while keeping the overall energy consumption low. This will be achieved by deploying highly specialized hardware accelerators and a reasonable resource management. First results show the viability of the methods in the distributed management of available