Message-ID: <45F0654D.8040800@mochima.com>
Date: Thu, 08 Mar 2007 14:34:37 -0500
From: Carlos Moreno <moreno_pg@mochima.com>
User-Agent: Mozilla Thunderbird 1.0.8-1.1.fc4 (X11/20060501)
MIME-Version: 1.0
To: PostgreSQL Performance <pgsql-performance@postgresql.org>
Subject: Re: strange performance regression between 7.4 and 8.1
References: <a728f9f90703011240r9657ac5ra7361fdfdc55c10f@mail.gmail.com>	
	<2D7248A6-38DB-4FE2-A570-6476D5985E85@event-s.net>	
	<a728f9f90703051856s2562f4a4g2fd0ba59f3b97609@mail.gmail.com>	
	<26FE4DE3-7896-4A8F-B26A-4FCDA1890078@event-s.net>	
	<a728f9f90703060725u56b902e7s64ede481536433de@mail.gmail.com>	
	<E1HOczX-0007ua-CZ@elasmtp-banded.atl.sa.earthlink.net>	
	<a728f9f90703061011m7d7165b5t70d0f17b6391b096@mail.gmail.com>
	<3cf983d0703080423x7d2e57d4w52d883f6812d3031@mail.gmail.com>
	<45F0571E.20501@modgraph-usa.com>
In-Reply-To: <45F0571E.20501@modgraph-usa.com>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit


>> I would just like to note here that this is an example of inefficient 
>> strategy.
>> [ ... ]
>
>
> Alex may have made the correct, rational choice, given the state of 
> accounting at most corporations.  Corporate accounting practices and 
> the budgetary process give different weights to cash and labor.  Labor 
> is fixed, and can be grossly wasted without (apparently) affecting the 
> quarterly bottom line.  Cash expenditures come directly off profits.
>
> It's shortsighted and irrational, but nearly 100% of corporations 
> operate this way.  You can waste a week of your time and nobody 
> complains, but spend a thousand dollars, and the company president is 
> breathing down your neck.
>
> When we answer a question on this forum, we need to understand that 
> the person who needs help may be under irrational, but real, 
> constraints, and offer appropriate advice.  Sure, it's good to fight 
> corporate stupidity, but sometimes you just want to get the system 
> back online.


Another thing --- which may or may not apply to Alex's case and to the
particular state of the thread, but it's still related and IMHO 
important to
take into account:

There may be other consrtaints that makes it impossible to even consider
a memory upgrade --- for example, us (our project).  We *rent* the servers
from a Web hoster (dedicated servers).  This particular hoster does not
even offer the possibility of upgrading the hardware --- 2GB of RAM,
take it r leave it.  Period.

In other cases, the memory upgrade has a *monthly* cost  (and quite
often I find it excessive --- granted, that may be just me).  So, $50 or
$100 per month *additional* expenses may be considerable.

Now, yet another thing that you (Craig) seem to be missing:  you're
simply putting the expense of all this time under the expenses column
in exchange for solving the particular problem --- gaining the insight
on the internals and performance tuning techniques for PG may well
be worth tens of thousands of dollars for his company in the future.

The "quick and dirty" solution is not giving a damn about knowledge
but to the ability to solve the problem at hand *now*, at whatever
"petty cash cost" because it looks more cost effective (when seen
from the non-irrational accounting point of view, that is) --- but isn't
going for the "quick and dirty" solution without learning anything
from the experience also shortsighted ???

Carlos
--