15.00 - Consumption of Disk Space by Populated and Empty Partitions - Teradata Database

Teradata Database Design

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

Consumption of Disk Space by Populated and Empty Partitions

With the large number of partitions that can be defined for a table or join index, it is very likely that a high percentage of those partitions are empty at any given time. For example, a table on a 200 AMP system that defines 100,000 combined partitions with 100 rows per unaligned (127.5 KB) data block and 100 data blocks per each combined partition per AMP has 200 billion rows. This is a relatively small number of combined partitions when you consider that the maximum for a table or join index is 9,223,372,036,854,775,807 combined partitions.

If each row were 100 bytes in length, the primary data alone consumes 20 petabytes of disk. That is 20 x 1015 bytes. It is highly unlikely that every combined partition would be populated. When you consider a multidimensional use of multilevel partitioning, you can easily deduce that not all combinations of dimension values actually occur.

In general, a populated combined partition should have either many data blocks per AMP or no data blocks. For the example proposed in the first paragraph, if there is actually only 200 gigabytes of data and each populated combined partition had 100 data blocks per AMP, about 99% of the combined partitions are empty.