In order to cater to a wide array of tasks such as DML, bulk loading, and querying, Vertica implements a storage model as shown in the following illustration. This model is the same on each Vertica node.
The Read Optimized Store (ROS) resides on a physical disk storage structure and is organized by projection. Compression is employed at the ROS level to ensure that the disk space occupied by projections is minimal.
Unlike the ROS, the Write Optimized Store (WOS) resides on primary memory, again organized by projection. The WOS stores and sorts the data by epoch. It doesn't compress the data at this level, ensuring high speed writes.
To learn more on how we can control data writing to WOS and ROS, refer to the Bulk Loading section.
Tuple Mover (TM) is a Vertica background process, which is responsible for moving data from primary memory (WOS) to physical disk (ROS). Apart from that, TM is also responsible for removing deleted data and coalescing small...