Efficient query processing in any data
management system typically relies on: (a) A profiling component that
gathers statistics used to evaluate possible execution plans, and (b) A
planning component that picks the plan with the best predicted
performance. I will first describe a range of modern data management
applications, including sensor data processing, continuous queries over
data streams, and query processing over web services, for which
traditional profiling and planning techniques are inadequate. I will
then focus on two specific contributions: a new planning technique for
query processing over web services, and a new profiling technique that
avoids the need for periodic, full scans of data. I will also summarize
my results for profiling and planning in other modern data management
scenarios, and I will outline directions for future work in this general
area.