Using Hash-Based Statistics

The partial_hashing parameter uses hashing on columns only when it produces histograms that are as accurate and of equivalent quality as those produced by sorting (low-domain cases), otherwise, the partial_hashing parameter uses sorting for high-domain cases.

Hashing may produce less accurate histograms than sorting in some cases (high-domain cases).

Although hash-based statistics do not require the tempdb disk space or procedure cache memory used by sorting, it may use a significant amount of tempdb buffer cache memory.

Large sorts used for the no_hashing parameter may scavenge statements in the statement cache and stored procedures to release procedure cache memory to support the sort.

max_resource_granularity limits the amount of tempdb buffer cache memory used for hashing or partial_hashing. It does not affect the amount of memory used by no_hashing parameter or sorting.

If you include the partial_hashing parameter, and a previous histogram on a column exists that indicates a high-domain column, the SAP ASE server assumes that sort-based statistics are required on this column. If no previous histogram exists on a column, then the SAP ASE server assumes this column is low-domain until the high-domain limit is reached.