Performance Effects of Duplicate NUPI Row Checks - Teradata Database

Teradata Database Design

Product
Teradata Database
Release Number
15.10
Language
English (United States)
Last Update
2018-10-06
Product Category
Software

Performance Effects of Duplicate NUPI Row Checks

Duplicate row checking can be a very I/O-intensive operation. The table in the topic “Duplicate NUPI Row Read I/O as a Function of Number of Rows Inserted” on page 393 illustrates the number of logical reads performed for the duplicate row checks required by a given number of duplicate NUPI row insertions under the following conditions:

  • The table is a SET table (no duplicate rows permitted).
  • There are no previous hash synonyms for the NUPI value being inserted.
  • The center column in the table indicates the number of logical reads required before the number of rows specified in the left column can be inserted into the table.

    The right column indicates the cumulative number of logical reads required before the number of rows listed in the left column can be inserted. The cumulants are determined from the following equation.

    More formally, the number of NUPI rows that must be read can be expressed as follows.

    where n = number of NUPI rows.