FILLFACTOR

Hi, The 'Informix Handbook' seems to indicate if more inserts are to be expected in the future, then the FILLFACTOR should be small to improve insertion performance. My questions: 1. How to determine that a fillfactor is (or not) optiomal ? 2. If I am going to change the FILLFACTOR in the onconfig for a running system, then can I just bounce IDS or I have to do something else ? 3. If the FILLFACTOR is 50, then I can understand that when a page is full, it will be split into 50/50. How does IDS split the page when the FILLFACTOR is, say, 60 or 80 ? Thanks, Thanh Sent via Deja.com http://www.deja.com/ Before you buy.

Thanh wrote: > > Hi, > > The 'Informix Handbook' seems to indicate if more inserts are to be > expected in the future, then the FILLFACTOR should be small to improve > insertion performance. My questions: > > 1. How to determine that a fillfactor is (or not) optiomal ? Guess. Estimate number of final rows versus current rows and use that ratio for FILLFACTOR. If the table is empty FILLFACTOR is noop. > 2. If I am going to change the FILLFACTOR in the onconfig for a running > system, then can I just bounce IDS or I have to do something else ? That's it, but read below: > 3. If the FILLFACTOR is 50, then I can understand that when a page > is full, it will be split into 50/50. How does IDS split the page > when the FILLFACTOR is, say, 60 or 80 ? NO! FILLFACTOR ONLY effects the fillness of newly created index nodes at CREATE time it does not effect the rebalancing of pages on the fly. When a node is split upon inserting a key it is always split into two equal nodes each 50% full. Art S. Kagel