Random distribution on fewer extents

You can configure the Netezza Performance Server system to create random distribution tables using a smaller number of extents to increase the number of user tables that can be created in the system. This feature is referred to as random chunk distribution.

The maximum number of non-empty tables on the system depends on the table row counts and widths and on the number of extents that the tables use. By default, the Netezza Performance Server system follows the storage rules in Table size and storage space, which distributes table rows randomly across data slices and allocates a minimum 3 MB extent on each data slice that holds at least one row of the table. For very small tables, this storage allocation process typically results in a larger allocation of space to accommodate new and updated rows for the table and to leverage the parallel query processing performance by dividing the work across multiple data slices, disks, and SPUs in the system.

If your users typically have many very small tables, such as those that actually require only one or a few extents of total storage, it might be beneficial to keep the records together in one or a few extents, instead of using additional extents. If you enable the random chunk distribution feature, the system is forced to store the records in fewer extents, the table allocation is smaller, and users can create more small tables. With this feature, the system can support up to 200,000 non-empty tables, or even more.

The random chunk distribution feature is typically used for tables that are very small and mostly static after they are created. Larger or more dynamic tables should use distribution keys (hash distributions) to ensure that their records are well distributed and that they take advantage of the Netezza Performance Server parallel performance benefits.