Approximate query on historical stream data

Abstract

We present a new Stream OLAP framework to approximately answer queries on historical stream data, in which each cell is extended from a single value to a synopsis structure. The cell synopses can be constructed by the existing well researched methods, including Fourier, DCT, Wavelet, PLA, etc. To implement the Cube aggregation operation, we develop algorithms that aggregate multiple lower level synopses into a single higher level synopsis for those synopsis methods. Our experiments provide comparison among all used synopsis methods, and confirm that the synopsis cells can be accurately aggregated to a higher level. © 2011 Springer-Verlag Berlin Heidelberg.