One component of this
project focuses on the issues surrounding the efficient processing
of declarative queries within sensor networks, looking in particular
at the ways in which traditional, relational query optimizers must
change to support the lossy, streaming data from sensornets. Query
optimization has the potential to dramatically reduce the energy
costs associated with sensor network data collection, offering dramatic
increases in network lifetime bandwidth usage.
The second component of the project focusses on data integration
issues. For sensor networks to become truly ubiquitous, it is important
that such networks be able to integrate with data sources and share
query processing with databases outside of the network, including
traditional relational databases, flat files, and even other, remote
sensor networks. To enable such integration, systems
must be enhanced to allow them to understand each other's capabilities
and offload query processing tasks when there are significant energy
or computational savings to be had. This project seeks to seamlessly
integrate sensor network query processing systems with other database
systems, be they conventional relational DBMSes or other sensornets.
To do this, we introduce a proxy to eliminate the need to make invasive
changes to pre-existing software and propose techniques based upon
an optimization framework that chooses to push down computation
when it will decrease the power consumption of the sensor network.