Framing the question: detecting and filling spatial-temporal windows

Abstract

We propose a new mechanism, which we term frames, for data-dependent windows. In contrast to traditional timestamp-based windows, frames represent just the boundary of a window and can be filled with data from secondary streams or historical data. Examples show how frames can be useful in network and sensor monitoring applications. We present frame definition and implementation in one dimension, discuss extension to multidimensional frames, and identify issues for further investigation.