Abstract

A sensor network is a set of data-source nodes (sensors) and computation nodes (processors) interconnected by a network. The computations that process the stream-data coming form from the sensors need to be mapped to the processors. This can be understood as an optimization problem with throughput as the objective function. However, there are some constraints to regard. Both computations and sensor network change under execution: applications are added and removed and nodes fail, are replaced, or added. Moreover, the sensor network is a limited resource, which research groups are competing for, i.e., the optimum throughput of the computations does not necessarily optimize the groups' utility. We present scheduling techniques which work under the above constraints.

The presentation (PDF)