Computational Indexes for Time Series

Esa Falkenroth

Dept. of Computer and Information Science, Linköping University, Sweden


A fast temporal indexing method for main memory databases is proposed. We consider storing time stamped data from simulation and sensing applications and the subsequent retrieval of the information given a time stamp. The indexing technique is based on super-linear equation solving methods and assumes that new data arrives with monotonically increasing time stamps. We also present some results of a performance measurement of the storage structure and the indexing technique. The time complexity is O(log(log(n))) on the average for typical simulation data.

Download the paper (49826 bytes).

Esa Falkenroth <>