A user-defined data type for the storage of time series data allowing efficient similarity screening

详细信息	查看全文 | 推荐本文 |

• 作者：Anatoly Sorokin^a ; ^b ; ^{lptolik@gmail.com} ; Gene Selkov^c ; Igor Goryanin^a
• 关键词：Time series ; Storage ; Indexing ; Database ; Experimental data
• 刊名：European Journal of Pharmaceutical Sciences
• 出版年：2012
• 期刊代码：15_09280987
• 类别：pha
• 出版时间：16 July, 2012
• 卷：46
• 期：4
• 页码：272-274
• 文件大小：210 K

摘要

The volume of the experimentally measured time series data is rapidly growing, while storage solutions offering better data types than simple arrays of numbers or opaque blobs for keeping series data are sorely lacking. A number of indexing methods have been proposed to provide efficient access to time series data, but none has so far been integrated into a tried-and-proven database system. To explore the possibility of such integration, we have developed a data type for time series storage in PostgreSQL, an object-relational database system, and equipped it with an access method based on SAX (Symbolic Aggregate approXimation). This new data type has been successfully tested in a database supporting a large-scale plant gene expression experiment, and it was additionally tested on a very large set of simulated time series data.