Re: FW: Sequential scan

<SNIP> > The reason is, bouncing up and down an index is hard work - quite a few > pages have to be read. If you are going for a few rows out of millions, then > of course this is where indexes really shine. Once the number of rows > selected via the index reaches a sufficiently large portion of the total > number of rows, you will reach a point where the index search is the same > speed as a TOTAL sequential search, and beyond that, it will become slower > than a seq. search. Somewhere around 10% to 50% of total rows is the > breakeven point, depending on the schema and data. The optimiser must think > this is going to happen. I'd say you don't have many pay periods in your > database yet. Just so people know, I have seen the breakeven point as low as .8% of the data ... <SNIP> Will Sent via Deja.com http://www.deja.com/