A Comparative Study of Row and Column Storage for Time Series Data

Over the past few decades, researchers have done massive research on data storage structures of relational databases. The existing research mainly focuses on the analysis and optimization of row and column storage structures in relational databases. It is discovered that row and column storage techniques are acceptable for relational database operations in a variety of goal and usage scenarios. However, with the generation of massive time series data, researchers ignore experimental analysis for specific storage structures for time series data in databases. In order to provide comprehensive verification, we compare and analyze the space and time consumed in the process of bulk loading and insertion, range query, and aggregate calculation of time series data under openGauss-based row and column storage. The purpose is to provide a storage design basis for extending time series data management capabilities in relational databases or developing time series databases. The results show that the choice of storage for time series should be based on the specific application scenarios.