These columns summarize the effects of prefetching data blocks on the file system. For more information, see “Data Block Prefetches Columns” in ResUsageSvpr Table Overview.
A prefetch is either a cylinder read operation or individual block reads operation. Either of these operations are generically called a prefetch.
When all cylinder slots are in use, the cylinder reads revert back to the original algorithm of a block-at-a-time read ahead. So the column FilePreKB is the sum of the size of data blocks logically read by either cylinder reads or data block pre-reads. This also applies to the physical pre-reads. FilePreReadKB includes both physical cylinder reads and single block pre-reads.
The number of data blocks that are pre-read at a time is controlled by the DBS Control performance parameter ReadAhead Count. The default is one block at a time pre-read.
If you enable cylinder reads, there will be extra sectors read in on cylinder reads. An accurate calculation of the wasted KB read by cylinder read is not possible since there are legitimate logical pre-reads that do not incur physical pre-reads.
For more information about cylinder reads, see Teradata Vantage™ - Database Administration, B035-1093.
|Column Name||Mode||Description||Data Type|
|FilePreKB||count||Sum of the sizes of data blocks logically loaded with data prefetches (for example, either cylinder reads or individual block reads).
For cylinder reads, this column does not include the disk sectors in between the loaded data blocks.
|FilePreReadKB||count||Size of the data prefetch (cylinder section or individual blocks being read) that is physically loaded from disk.
For cylinder reads, this column includes the disk sectors in between the loaded data blocks.
|FilePreReads||count||Number of times a data prefetch was physically performed either as a cylinder read or individual blocks read.||FLOAT|
|FilePres||count||Total number of times a logical data prefetch was performed (either as a cylinder read or individual block reads).||FLOAT|