ICB-Index: A New Indexing Technique for Continuous Time Sequences

Various application domains require databases to store time sequences. Very often time sequences describe some continuous processes at discrete time points. Many applications require queries to take into consideration not only explicit values of time sequences, but also the values of the processes represented by them (these values can be derived from explicit values by user-defined interpolation functions). For example, a user of industrial process control system may ask the following query: "Find those time intervals during which specified physical value, represented by a series of measurements, was greater than given limit value". We show that conventional secondary indexes are not suitable to support such queries. We also investigate the properties of IP-index – the first index structure supporting queries on time sequences taking into account the interpolation (so-called "queries on continuous time sequences"). We show that IP-index improves the performance of such queries, but its size is enormously big for many real-life sequences. This fact makes it nearly impossible to use IP-index in some application domains. In this paper we present a new indexing technique to support queries on continuous time sequences – ICB-index. ICB-index makes the performance of such queries as high as IP-index does, but it requires substantially less space than IP-index. The effectiveness of ICB-index is verified by experiments on sensor-generated time sequences from a power plant.