Re: Clustered indexes - again

Almost by definition of 'look-up table', it won't matter in the slightest whether the index is clustered or not. A lookup table will be accessed semi-randomly. A clustered index gives performance benefits when reading sequentially through a table in the order of the index because the data rows are in the same sequence as the index, so every page read to fetch one row has effectively pre-fetched the remaining rows on the page. A side effect of that is that if you have large rows (say 1 or 2 to a page, or even 1 row using several pages), then the performance benefit from a clustered index is probably smaller than if you have a larger number of rows (say 20 or 30) to the page. The concept of pages applies to SE as well as OnLine, though the pages in SE are the ones allocated by the Unix file system (on Unix, of course). Yours, Jonathan Leffler (johnl@informix.com) #include <disclaimer.h> >Date: Wed, 13 Mar 1996 17:03:59 -0500 (EST) >From: Nick Nobbe <nnob@loc.gov> >X-Informix-List-Id: <list.9034> > >[...] > >What if you have a lookup table? Wouldn't you want to cluster index on >that to move things along?