Techniques for automatic data placement with compression and columnar storage

摘要

For automatic data placement of database data, a plurality of access-tracking data is maintained. The plurality of access-tracking data respectively corresponds to a plurality of data rows that are managed by a database server. While the database server is executing normally, it is automatically determined whether a data row, which is stored in first one or more data blocks, has been recently accessed based on the access-tracking data that corresponds to that data row. After determining that the data row has been recently accessed, the data row is automatically moved from the first one or more data blocks to one or more hot data blocks that are designated for storing those data rows, from the plurality of data rows, that have been recently accessed.

主权项

1. A computer-implemented method comprising:
maintaining a plurality of access-tracking data about accesses to a plurality of data rows in a database that is managed by a database server, said plurality of access-tracking data including first access-tracking data about accesses to a first data row of said plurality of data rows, said database comprising multiple data blocks that store said plurality of data rows, said multiple data blocks comprising a plurality of hot data blocks and a plurality of cold data blocks, where the plurality of hot data blocks are designated for storage of a type of storage that is faster than a type of storage designated for said plurality of cold data blocks; while the database server is executing database transactions that change said database, performing steps comprising:
based on said first access-tracking data, determining to move said first data row from a cold data block of said plurality of cold data blocks to a hot data block of said plurality of hot data blocks;in response to determining to move said first data row from a cold data block of said cold data blocks to a hot data block of said hot data blocks, removing said first data row from first one or more cold data blocks of said plurality of cold data blocks and storing the first data row into first one or more hot data blocks of said plurality of hot data blocks; and wherein the method is performed by one or more computing devices.