Experimenting with segmentation and non-segmentation methods for storing temporal data

摘要

We have defined a new temporal query language, SQL~*T. SQL~*T is an intuitive and straightforward extension of SQL2. SQL~*T will be implemented on a conventional relational DBMS. We must decide the proper data management method for implementation of SQL~*T. In this paper, we compared and analyzed three data management methods for temporal data: [P][C][F]-method, [PCF]-method, and [C][PF]-method. In order to compare the response time of the three data management methods for temporal data, we varied rates: moving rate of current and future data at a granularity level (second, minute, and hour), respectively. The key concept of [P][C][F] and [C][PF]-method is that Searching and Moving Process is triggered at a granularity level. The results of this experiment are as follows: [PCF]-method had the best response time among the three data management methods when granularity level is second. [P][C][F]-method had the best performance among the three data management methods when granularity level is hour. The performance gap between [P][C][F]-method and [C][PF]-method was 5percent when granularity level is minute or hour. When granularity level was second or minute, [P][C][F] and [C][PF]-method were very sensitvie to data moving rate.