The point Gentlemen, was that Good Architecture is King. That's what I was trying to emphasize by calling proper DB architecture step 0. All other things being equal (and they usually aren't, this sort of stuff is _very_ context dependent), the more of your critical schema that you can fit into RAM during normal operation the better.
...and it all starts with proper DB design. Otherwise, you are quite right in stating that you risk wasting time, effort, and HW. Ron -----Original Message----- From: Frank Wiles <[EMAIL PROTECTED]> Sent: Nov 9, 2005 6:53 PM To: Alvaro Herrera <[EMAIL PROTECTED]> Cc: [EMAIL PROTECTED], [EMAIL PROTECTED], pgsql-performance@postgresql.org Subject: Re: [PERFORM] Some help on buffers and other performance tricks On Wed, 9 Nov 2005 20:07:52 -0300 Alvaro Herrera <[EMAIL PROTECTED]> wrote: > IMHO you should really be examining your queries _before_ you do any > investment in hardware, because later those may prove unnecessary. It all really depends on what you're doing. For some of the systems I run, 4 GBs of RAM is *WAY* overkill, RAID 1+0 is overkill, etc. In general I would slightly change the "order of operations" from: 1) Buy tons of RAM 2) Buy lots of disk I/O 3) Tune your conf 4) Examine your queries to 1) Tune your conf 2) Spend a few minutes examining your queries 3) Buy as much RAM as you can afford 4) Buy as much disk I/O as you can 5) Do in depth tuning of your queries/conf Personally I avoid planning my schema around my performance at the start. I just try to represent the data in a sensible, normalized way. While I'm sure I sub-consciously make decisions based on performance considerations early on, I don't do any major schema overhauls until I find I can't get the performance I need via tuning. Obviously there are systems/datasets/quantities where this won't always work out best, but for the majority of systems out there complicating your schema, maxing your hardware, and THEN tuning is IMHO the wrong approach. --------------------------------- Frank Wiles <[EMAIL PROTECTED]> http://www.wiles.org --------------------------------- ---------------------------(end of broadcast)--------------------------- TIP 5: don't forget to increase your free space map settings ---------------------------(end of broadcast)--------------------------- TIP 9: In versions below 8.0, the planner will ignore your desire to choose an index scan if your joining column's datatypes do not match
