Re: Sysindexes - column defs

In article <500q6g$k82@nic.global-one.no>, Nils.Myklebust@ccmail.telemax.no (Nils Myklebust) says: > > >When you ask like this there are too many variabls left for anyone to >guess. >I wouldn't be too much conserned about the meaning of the sysindexes >columns. They sound like they have something to do with the size of >the B-tree and whether the index is clustered and/or unique. > >As to your speed problem. Same search using the same index i take as >meaning you have checked your sqlexplain.out file. (If not you can't >have any idea at all what index(s), if any, is used.) Yes, I have checked the sqexplain for both databases. They are exactly alike except for the number of estimated rows returned and cost. The index chosen has "street" as the first field. >Do you select one or many rows? What is the timing difference? Have >you dropped and recreated the indexes? Have you checked number of >extents and fragmentation of them? Have you done anything about it? >(Unload/load or alter index to cluster.) What else do you see? The slow database is 1 min, fast database 7 secs. Both dbase has been recently defrag'd so there is only 1 extent, they do not grow. Update statistics has been run. Some differences are (from memory): slow dbase fast dbase time for search 1 min 7 sec total size 560,000+ 190,000+ num unique keys 2400 5400 (street name) num rows selected 240 140 TIA, Steve sgoldber@landata.com