Difference B/W Generate Statistics & Generate Express Statistics in Netzza

        In 4.6 and 5.0 (and all newer releases), generate statistics automatically determines if it will do express or full statistics based off the size of the table. So if the table is large Netezza will gather a statistical estimate instead of gathering full statistics.

        Generate express statistics was deprecated due to the general netezza philosophy that things should be simple. On Oracle you have 57 different options you can use to gather statistics. It takes many hours of study time to even understand what all that stuff does.

Netezza now has 1 option "generate statistics on < >" that does the right thing automatically.

        The secondary reason is that gathering full statistics on a large table is generally useless anyway. Sometimes it helps and sometimes it hurts, but it is always very expensive. JIT stats gives a much more reliable estimate of the data that will be picked up by a query from a large table based off the constraints in the query itself instead of relying on out dated statistics.

No comments:

Post a Comment