SCMAT: A Mechanism Presuming SCMs to Efficiently Enable Both OLAP and OLTP

Many commercial DBMSs based on a column-oriented storage method have been used to analyze large-scale data. However, the column-oriented DBMS is difficult to use for processing big-data analysis updated in real time, because the column-oriented storage performs inefficient OLTP when processing row-oriented updates. Therefore, we attempted to enable the column-oriented DBMS to efficiently process OLTP for performing big data analysis. In this paper, we propose a DBMS architecture by focusing on storage class memory (SCM) such as STT-MRAM, PRAM, and ReRAM of new storage devices used in future computing. Our approach assumed that SCMs have not yet been considered, we propose a TiD-based update index for modifying the column data using SCMs in real time. Moreover, the DBMS mechanism we consider is able to identify a row of the column data such that we efficiently use a materialization method for aggregation in OLAP in which users perform a similar OLAP query for data analysis.