pg.ddx.io  pgsql-performance@postgresql.org mailing list archive  
help / color / mirror / Atom feed
From: Jochem van Dieten <jochemd@oli.tudelft.nl>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Manfred Koizar <mkoi-pg@aon.at>
Cc: pgsql-performance@postgresql.org
Subject: Re: planner/optimizer question
Date: Sat, 01 May 2004 13:18:04 +0200
Message-ID: <4093876C.2060409@oli.tudelft.nl> (raw)
In-Reply-To: <8881.1083374379@sss.pgh.pa.us>
References: <pvc290584alms4s6hurvsh5ra09ki7se1s@email.aon.at>
	<409290F0.10509@oli.tudelft.nl>
	<toe590tgu39ilahg3p1os2mjlgf4iss582@email.aon.at>
	<8881.1083374379@sss.pgh.pa.us>

Tom Lane wrote:
> Manfred Koizar <mkoi-pg@aon.at> writes:
>> 
>> Yes, the visible-to-all flag would be set as a by-product of an index
>> scan, if the heap tuple is found to be visible to all active
>> transactions.  This update is non-critical
> 
> Oh really?  I think you need to think harder about the transition
> conditions.
> 
> Dead-to-all is reasonably safe to treat as a hint bit because *it does
> not ever need to be undone*.  Visible-to-all does not have that
> property.

Yes, really :-)

When a tuple is inserted the visible-to-all flag is set to false. 
The effect of this is that every index scan that finds this tuple 
has to visit the heap to verify visibility. If it turns out the 
tuple is not only visible to the current transaction, but to all 
current transactions, the visible-to-all flag can be set to true.
This is non-critical, because if it is set to false scans will 
not miss the tuple, they will just visit the heap to verify 
visibility.

The moment the heap tuple is updated/deleted the visible-to-all 
flag needs to be set to false again in all indexes. This is 
critical, and the I/O and (dead)lock costs of unsetting the 
visible-to-all flag are unknown and might be big enough to ofset 
any advantage on the selects.

But I believe that for applications with a "load, select, drop" 
usage pattern (warehouses, archives etc.) having this 
visible-to-all flag would be a clear winner.

Jochem

-- 
I don't get it
immigrants don't work
and steal our jobs
     - Loesje




view thread (37+ messages)  latest in thread

Message-ID: <4093876C.2060409@oli.tudelft.nl>
Permalink:  ../4093876C.2060409@oli.tudelft.nl/
Also on:    postgresql.org/message-id/4093876C.2060409@oli.tudelft.nl

 · 

reply

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Reply to all the recipients using the --to and --cc options:
  reply via email

  To: pgsql-performance@postgresql.org
  Cc: jochemd@oli.tudelft.nl, tgl@sss.pgh.pa.us, mkoi-pg@aon.at
  Subject: Re: planner/optimizer question
  In-Reply-To: <4093876C.2060409@oli.tudelft.nl>

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

This inbox is served by DDX for PostgreSQL; see mirroring instructions
for how to clone and mirror all data and code used for this inbox