Re: docco on external storage?

Lists: pgsql-hackers
From: Andrew Dunstan <andrew(at)dunslane(dot)net>
To: Postgresql Hackers <pgsql-hackers(at)postgresql(dot)org>
Subject: docco on external storage?
Date: 2003-11-03 17:53:46
Message-ID: 3FA6962A.1060706@dunslane.net
Views: Raw Message | Whole Thread | Download mbox | Resend email
Lists: pgsql-hackers


Perhaps I am missing something, but the documentation on external
storage appears remarkably sparse. It says it is better for substring
searches, but otherwise doesn't say very much.

I have a table with two modest-length bytea fields (around 9K and 2K
respectively) of already compressed data. I was looking for a clue as to
whether or not these should use external storage - I suspect they should
but I didn't see anything that gave me a definite answer.

I'd prepare doc updates on this if they were required but I don't have
the requisite knowledge, nor time right now to go trawling endlessly to
acquire it.

cheers

andrew


From: Joe Conway <mail(at)joeconway(dot)com>
To: Andrew Dunstan <andrew(at)dunslane(dot)net>
Cc: Postgresql Hackers <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: docco on external storage?
Date: 2003-11-04 02:50:33
Message-ID: 3FA713F9.6020407@joeconway.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Lists: pgsql-hackers

Andrew Dunstan wrote:
> I have a table with two modest-length bytea fields (around 9K and 2K
> respectively) of already compressed data. I was looking for a clue as to
> whether or not these should use external storage - I suspect they should
> but I didn't see anything that gave me a definite answer.

From src/include/access/tuptoaster.h:
-------------------------------------
/*
* These symbols control toaster activation. If a tuple is larger than
* TOAST_TUPLE_THRESHOLD, we will try to toast it down to no more than
* TOAST_TUPLE_TARGET bytes. Both numbers include all tuple header and
* alignment-padding overhead.
*
* The numbers need not be the same, though they currently are.
*/
#define TOAST_TUPLE_THRESHOLD (MaxTupleSize / 4)
#define TOAST_TUPLE_TARGET (MaxTupleSize / 4)

And from src/include/access/htup.h:
-----------------------------------
/*
* MaxTupleSize is the maximum allowed size of a tuple, including header
* and MAXALIGN alignment padding. Basically it's BLCKSZ minus the
* other stuff that has to be on a disk page. The "other stuff"
* includes access-method- dependent "special space", which we assume
* will be no more than MaxSpecialSpace bytes (currently, on heap pages
* it's actually zero).
*
* NOTE: we do not need to count an ItemId for the tuple because
* sizeof(PageHeaderData) includes the first ItemId on the page.
*/
#define MaxSpecialSpace 32
#define MaxTupleSize \
(BLCKSZ - MAXALIGN(sizeof(PageHeaderData) + MaxSpecialSpace))

And from src/include/pg_config_manual.h:
----------------------------------------
#define BLCKSZ 8192

So, it looks like the threshold for external storage is somewhere around
2000 bytes.

Joe


From: "Andrew Dunstan" <andrew(at)dunslane(dot)net>
To: "Postgresql Hackers" <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: docco on external storage?
Date: 2003-11-04 12:18:45
Message-ID: 005301c3a2cd$cb5d4a70$6401a8c0@DUNSLANE
Views: Raw Message | Whole Thread | Download mbox | Resend email
Lists: pgsql-hackers


Thanks. We don't really expect DBAs to go hunting through the source code,
though, do we?

cheers

andrew

----- Original Message -----
From: "Joe Conway" <mail(at)joeconway(dot)com>
To: "Andrew Dunstan" <andrew(at)dunslane(dot)net>
Cc: "Postgresql Hackers" <pgsql-hackers(at)postgresql(dot)org>
Sent: Monday, November 03, 2003 9:50 PM
Subject: Re: [HACKERS] docco on external storage?

> Andrew Dunstan wrote:
> > I have a table with two modest-length bytea fields (around 9K and 2K
> > respectively) of already compressed data. I was looking for a clue as to
> > whether or not these should use external storage - I suspect they should
> > but I didn't see anything that gave me a definite answer.
>
> From src/include/access/tuptoaster.h:
> -------------------------------------
> /*
> * These symbols control toaster activation. If a tuple is larger than
> * TOAST_TUPLE_THRESHOLD, we will try to toast it down to no more than
> * TOAST_TUPLE_TARGET bytes. Both numbers include all tuple header and
> * alignment-padding overhead.
> *
> * The numbers need not be the same, though they currently are.
> */
> #define TOAST_TUPLE_THRESHOLD (MaxTupleSize / 4)
> #define TOAST_TUPLE_TARGET (MaxTupleSize / 4)
>
>
> And from src/include/access/htup.h:
> -----------------------------------
> /*
> * MaxTupleSize is the maximum allowed size of a tuple, including header
> * and MAXALIGN alignment padding. Basically it's BLCKSZ minus the
> * other stuff that has to be on a disk page. The "other stuff"
> * includes access-method- dependent "special space", which we assume
> * will be no more than MaxSpecialSpace bytes (currently, on heap pages
> * it's actually zero).
> *
> * NOTE: we do not need to count an ItemId for the tuple because
> * sizeof(PageHeaderData) includes the first ItemId on the page.
> */
> #define MaxSpecialSpace 32
> #define MaxTupleSize \
> (BLCKSZ - MAXALIGN(sizeof(PageHeaderData) + MaxSpecialSpace))
>
>
> And from src/include/pg_config_manual.h:
> ----------------------------------------
> #define BLCKSZ 8192
>
>
> So, it looks like the threshold for external storage is somewhere around
> 2000 bytes.
>
> Joe
>
>
> ---------------------------(end of broadcast)---------------------------
> TIP 9: the planner will ignore your desire to choose an index scan if your
> joining column's datatypes do not match