Re: [GENERAL] general question on two-partition table

2009-07-28 Thread Stephen Frost
* Janet Jacobsen (jsjacob...@lbl.gov) wrote: > I looked at the documentation for partitions - it is the case, right, that I > have to create the master table and the two partition tables (depending > on the value of rbscore) and then copy the records from the existing > table into the two partition

Re: [GENERAL] general question on two-partition table

2009-07-28 Thread Janet Jacobsen
Hi. Thanks for your reply. I looked at the documentation for partitions - it is the case, right, that I have to create the master table and the two partition tables (depending on the value of rbscore) and then copy the records from the existing table into the two partitions? Stephen Frost wrot

Re: [GENERAL] general question on two-partition table

2009-07-27 Thread Stephen Frost
* Janet Jacobsen (jsjacob...@lbl.gov) wrote: > If they are going to spend 95% of their time querying the > records that meet the 'good' criteria, what are the good > strategies for ensuring good performance for those queries? > (1) Should I partition the table into two partitions based on > the val

Re: [GENERAL] general question on two-partition table

2009-07-27 Thread David Wilson
On Mon, Jul 27, 2009 at 8:24 PM, Greg Stark wrote: > I think it would be even more interesting to have partial indexes -- > ie specified with "WHERE rbscore < cutoff". Yes- that's what I actually meant. Word got scrambled between brain and fingers... -- - David T. Wilson david.t.wil...@gmail.co

Re: [GENERAL] general question on two-partition table

2009-07-27 Thread Greg Stark
On Tue, Jul 28, 2009 at 1:08 AM, David Wilson wrote: > On Mon, Jul 27, 2009 at 7:52 PM, Janet Jacobsen wrote: > >> Can you suggest other strategies? > > Something that might be easier to play with is to create a (or > several, to speed up other queries) functional index on the comparison > between

Re: [GENERAL] general question on two-partition table

2009-07-27 Thread David Wilson
On Mon, Jul 27, 2009 at 7:52 PM, Janet Jacobsen wrote: > Can you suggest other strategies? Something that might be easier to play with is to create a (or several, to speed up other queries) functional index on the comparison between rbscore and the cutoff. It won't buy you anything on seq scans,

[GENERAL] general question on two-partition table

2009-07-27 Thread Janet Jacobsen
Hi. We have a table with 30 M records that is growing by about 100 K records per day. The experimentalists, whose data are in the table, have decided that they will focus on the records for which the value of one field, rbscore, is greater than a cut-off. However, they want to continue to store a