Integrity Checking Using a Checksum
Disk I/O and File System Data Structures
Teradata Database disk I/Os read from and write to the following data and file structures:
The file system first folds each sampled 64‑bit check data segment into a 32‑bit chunk using a barrel shift method that maximizes the minimum distance between the folded bytes to optimize the likelihood of detecting bursty corruption of adjacent bytes while conserving storage space in the cylinder index. It then computes the checksum by XORing the individual sample set together. See the illustration of how this XORing to compute a checksum is done in the topic “Creating and Verifying File System Checksums” on page 668.
Reads and writes to file system structures are not made in smaller units than the file system structures listed. Because of this, a full checksum of the data in units of a DB and CI is sufficient to verify its integrity when a read completes on one of these structures.