A data block is a disk-resident structure that contains one or more rows from the same table and is the smallest I/O unit for the file system. Data blocks are stored in physical disk sectors or segments, which are grouped in cylinders.
Vantage employs Cylinder Read to enable full-file scan operations to run more efficiently by reading a list of cylinder-resident data blocks with a single I/O operation. This reduces I/O overhead from once per data block to once per cylinder. This can significantly reduce the time it takes to do a full-table scan of large tables. During installation, Cylinder Read is enabled by default.
- Full-table scan operations under conditions such as:
- Large select
- Merge INSERT ... SELECT
- Merge delete
- Sum, average, minimum, maximum, and count aggregates
- Joins operations that involve many data blocks, such as merge or product joins.
Cylinder reads are not invoked when there is row partitioning elimination or column partitioning elimination.
For more information, see Managing I/O with Cylinder Read.