Re: [PERFORM] number of rows estimation for bit-AND operation

2009-08-21 Thread Slava Moudry
Hi, Sorry I don't understand how the numbers came so low. If you assume that 8 boolean fields take 1 byte each, so for 100M table it will be 800M bytes. How did your table fit in 3560616 bytes ? Using postgres 8.4.0 on Linux x64: create table staging.tmp_t1(a1 boolean, a2 boolean, a3 boolean, a

Re: [PERFORM] Number of tables

2009-08-21 Thread Jerry Champlin
I think this requirement can be lumped into the category of "right hammer, right nail" instead of the "one hammer, all nails" category. There are many memory only or disk backed memory based key value stores which meet your requirements like Reddis and memcached. -Jerry Jerry Champlin|Abs

Re: [PERFORM] number of rows estimation for bit-AND operation

2009-08-21 Thread Alvaro Herrera
Robert Haas escribió: > Scott, did you check whether a toast table got created here and what > the size of it was? A table with only bool columns (and, say, one int8 column) would not have a toast table. Only varlena columns produce toast tables. -- Alvaro Herrera

Re: [PERFORM] number of rows estimation for bit-AND operation

2009-08-21 Thread Robert Haas
On Thu, Aug 20, 2009 at 9:58 PM, Scott Marlowe wrote: > On Thu, Aug 20, 2009 at 7:32 PM, Scott Marlowe wrote: >> 2009/8/20 Slava Moudry : >>> Hi, >>> Yes, I thought about putting the bit-flags in separate fields. >>> Unfortunately - I expect to have quite a lot of these and space is an issue >>> w

[PERFORM] Performance regression between 8.3 and 8.4 on heavy text indexing

2009-08-21 Thread Gaël Le Mignot
Hello, We are using PostgreSQL to index a huge collection (570 000) of articles for a french daily newspaper (Libération). We use massively the full text search feature. I attach to this mail the schema of the database we use. Overall, we have very interesting performances, except in a few cas