Efficient interval indexing for content-based subscription e-commerce and e-service

Abstract

This paper presents a dynamic interval index for fast event matching against a large number of predicate intervals specified by content-based subscriptions. A set of virtual construct intervals (VCIs) are predefined, with a unique ID and an associated ID list. Each predicate interval is decomposed into one or more VCIs, which become activated by the predicate. The predicate ID is then inserted into the ID lists associated with the decomposed VCIs. To facilitate fast search, we start with a bitmap vector to indicate the activation of VCIs that cover an attribute value. Then, we study various techniques to reduce the storage cost, including logarithmic construct intervals (LCI), which reduce the total number of VCIs, bitmap clipping which prunes certain positions of a bitmap vector, and bitmap virtualization, which eliminates the bitmap. Simulations are conducted to evaluate and compare these techniques