15.00 - Computing the Break-Even Percentage - Teradata Database

Teradata Database Design

prodname
Teradata Database
vrm_release
15.00
category
User Guide
featnum
B035-1094-015K

Computing the Break-Even Percentage

Another way to look at breaking even is to compute the percentage of rows in the table that must be compressed in order to realize no net savings. When you can value‑compress a higher percentage of rows than this through compression (expressed as a higher value for the break‑even percentage), then you should compress the value.

Use the following equation to compute the break-even percentage for multi-value compression on a column.

Equation 1: Break-Even Percentage

where:

 

Variable name …

Specifies …

table_header_space

  • the number of table header bytes used by the compressed column for aligned row format systems.
  • the number of table header bytes used by the compressed column + 2 for packed64 format systems.
  • This value indicates the number of additional bytes required in the table header to compress the specified value, nulls, or both.

    fallback_factor

    whether fallback is enabled for the table.

  • 1 = fallback not enabled
  • 2 = fallback enabled
  • row_header_space

    the number of presence octets (see “Equation 9: Additional Presence Octets” on page 728) in the table.

    This value indicates the number of additional bytes required in the row header to compress the specified value, nulls, or both.

    cardinality

    the number of rows in the table.

    field_size

    the width of the compressed column in bytes.

    field_sizecol

    the number of table header bytes used by the compressed column + 1.

    Field size, fallback, table header space, row header space, and cardinality are all factors in this determination, so the break-even point is different depending on the demographics of the case being evaluated.