CEP and Storage – Where Does All the Data Go?
Many CEP systems, regardless of the method of expressing the processing, may fundamentally be thought of as a directed graph of relational primitives (or stream operators). If one arranges the graph from top to bottom (in terms of data flow) one can think of data entering the top in the most raw form, and as the data works its way down the graph, being refined though further stages of processing. A simple example of this notion is given in the following graphical representation of a market feed driven price/positions model.
The input data in this simple example enters into the Prices and Book streams, while the other streams, AveragePrices, IndiviudualPositions and BookPositions are all computed based on the input data.
The question posed in this paper is, where is the data stored for these streams, and what options for storage are there?