Wavescope: A Data Management System for Signals
                          Sam Madden, MIT

Sensors capable of sensing phenomena at high data rates on the order
of tens to hundreds of thousands of samples per second are now widely
deployed in many industrial, civil engineering, scientific,
networking, and medical applications.  In aggregate, these sensors
easily generate several million samples per second that must be
processed within milliseconds or seconds.  The required processing
typically involves a combination of both event stream and signal
processing operations.

In this talk, I will describe the architecture and implementation of
WaveScope, a system that combines these two different classes of
processing functions. WaveScope aims to improve both programmer
productivity, by making it easy to develop user-defined processing
functions, and achieves high performance, processing several million
samples per second on a standard PC.  It uses three key ideas to meet
its goals: a new basic data type, the signal segment, to efficiently
manipulate isochronous collections of sensor samples; a new language,
WaveScript, that makes it easy to write user-defined functions that
combine event stream processing and signal processing logic; and an
efficient memory manager and novel scheduler for high performance
execution.  Our experiments over three real-world
applications---acoustic monitoring, pipeline leak detection, and
network monitoring---suggest that WaveScope is a viable platform for
an emerging class of very high rate streaming applications.