Tom Lane writes:
Tatsuo Ishii [EMAIL PROTECTED] writes:
string_to_array() consumes too much memory. For example, to make
~70k array elements, string_to_array seems to eat several Gig bytes
of memory.
I'd argue that the problem comes from enlarging the work arrays only
64 elements at a time in
On Mon, Nov 06, 2006 at 09:50:53PM +, Simon Riggs wrote:
- There are specific issues with the optimizer's ability to understand
dead row numbers, which can in some cases lead to SeqScan plans that are
inappropriate when tables grow because of updates. This is a red-herring
that can lead
Heikki Linnakangas wrote:
Andrew Dunstan wrote:
Marc G. Fournier wrote:
As a result of there being two *known* outstanding bugs, we have
just bundled up a Beta3, to allow for testing of the recent patch
concerning WAL replay ...
What are the bugs?
AFAIK:
1. Tuple freezing and hint
On Thu, 2006-11-09 at 04:09 -0500, Andrew Sullivan wrote:
On Mon, Nov 06, 2006 at 09:50:53PM +, Simon Riggs wrote:
- There are specific issues with the optimizer's ability to understand
dead row numbers, which can in some cases lead to SeqScan plans that are
inappropriate when tables
Design Overview of HOT Updates
--
The objective is to increase the speed of the UPDATE case, while
minimizing the overall negative effects of the UPDATE. We refer to the
general requirement as *Frequent Update Optimization*, though this
design proposal is for Heap
Nice idea, just one question:
On Thu, Nov 09, 2006 at 05:13:16PM +, Simon Riggs wrote:
Behavioural Characteristics
---
With HOT, it is easily possible that the chain of prior versions spans
many blocks. The chain always starts with the block of the root tuple
On Thu, 2006-11-09 at 18:49 +0100, Martijn van Oosterhout wrote:
Nice idea, just one question:
It seems to me that bitmap index scans will get these same
characteristics also, right? The bitmap scan will have to follow the
chain of any possibly matching tuple in any of the blocks that are in
On Tue, 2006-11-07 at 15:00 +1300, Mark Kirkwood wrote:
Simon Riggs wrote:
EnterpriseDB has been running a research project to improve the
performance of heavily updated tables. We have a number of approaches
prototyped and we'd like to discuss the best of these now on -hackers
for
Simon,
If we perform an update that meets the HOT criteria then we put the
new version into the overflow relation; we describe this as a HOT
UPDATE. If we perform an update that does not meet the criteria, then we
carry on with the existing/old MVCC behaviour; we describe this as a
non-HOT
Quoth [EMAIL PROTECTED] (Simon Riggs):
On Tue, 2006-11-07 at 15:00 +1300, Mark Kirkwood wrote:
Simon Riggs wrote:
EnterpriseDB has been running a research project to improve the
performance of heavily updated tables. We have a number of approaches
prototyped and we'd like to discuss the
Tom Lane wrote:
Andrew Dunstan [EMAIL PROTECTED] writes:
Anyway, it is probably not expected by many users that loading a module
in plperlu makes it available to plperl - I was slightly surprised
myself to see it work and I am probably more aware than most of perl and
plperl subtleties.
I
On Thu, 2006-11-09 at 13:21 -0800, Josh Berkus wrote:
Simon,
If we perform an update that meets the HOT criteria then we put the
new version into the overflow relation; we describe this as a HOT
UPDATE. If we perform an update that does not meet the criteria, then we
carry on with the
Simon Riggs [EMAIL PROTECTED] writes:
...and of course it would be good if LISTEN/NOTIFY were able to use this
concept also, to help Slony along also.
LISTEN/NOTIFY are overdue to be rewritten to not use a table at all,
so I'm not particularly worried about whether this idea is applicable
to
Simon Riggs [EMAIL PROTECTED] writes:
As more UPDATEs take place these tuple chains would grow, making
locating the latest tuple take progressively longer.
This is the part that bothers me --- particularly the random-access
nature of the search. I wonder whether you couldn't do something
Tom,
Actually, you omitted to mention the locking aspects of moving tuples
around --- exactly how are you going to make that work without breaking
concurrent scans?
I believe that's the unsolved technical issue in the prototype, unless
Pavan has solved it in the last two weeks. Pavan?
--
It occurred to me to run the regression tests type_sanity and opr_sanity
over the contrib/isn code. (The easy way to do this is to copy the isn
install script into the regress/sql directory and then add it to
serial_schedule just before those two tests.) It turned up a couple
of moderately
Mark Dilger wrote:
Tom Lane wrote:
Andrew Dunstan [EMAIL PROTECTED] writes:
Anyway, it is probably not expected by many users that loading a
module
in plperlu makes it available to plperl - I was slightly surprised
myself to see it work and I am probably more aware than most of perl
and
On 11/10/06, Josh Berkus josh@agliodbs.com wrote:
Tom, Actually, you omitted to mention the locking aspects of moving tuples around --- exactly how are you going to make that work without breaking concurrent scans?I believe that's the unsolved technical issue in the prototype, unless
Pavan has
Pavan Deolasee [EMAIL PROTECTED] writes:
On 11/10/06, Josh Berkus josh@agliodbs.com wrote:
I believe that's the unsolved technical issue in the prototype, unless
Pavan has solved it in the last two weeks. Pavan?
When an overflow tuple is copied back to the main heap, the overflow tuple
is
On 11/10/06, Tom Lane [EMAIL PROTECTED] wrote:
Pavan Deolasee [EMAIL PROTECTED] writes: On 11/10/06, Josh Berkus
josh@agliodbs.com wrote:
I believe that's the unsolved technical issue in the prototype, unless Pavan has solved it in the last two weeks. Pavan? When an overflow tuple is copied
Ühel kenal päeval, N, 2006-11-09 kell 18:28, kirjutas Tom Lane:
Simon Riggs [EMAIL PROTECTED] writes:
As more UPDATEs take place these tuple chains would grow, making
locating the latest tuple take progressively longer.
This is the part that bothers me --- particularly the random-access
I was trying to compile 8.2beta3 on openbsd, and ran into an interesting
issue. My account on the particular openbsd box has some restrictive
ulimit settings, so I don't have a lot of memory to work with. I was
getting an out of memory issue linking postgres, while I did not before.
I figured
Pavan Deolasee [EMAIL PROTECTED] writes:
On 11/10/06, Tom Lane [EMAIL PROTECTED] wrote:
(2) Isn't this full of race conditions?
I agree, there could be race conditions. But IMO we can handle those.
Doubtless you can prevent races by introducing a bunch of additional
locking. The question
Ühel kenal päeval, R, 2006-11-10 kell 09:06, kirjutas Hannu Krosing:
Ühel kenal päeval, N, 2006-11-09 kell 18:28, kirjutas Tom Lane:
Simon Riggs [EMAIL PROTECTED] writes:
As more UPDATEs take place these tuple chains would grow, making
locating the latest tuple take progressively longer.
Jeremy Drake [EMAIL PROTECTED] writes:
I figured out that the -g flag was being surreptitiously added to my
CFLAGS. It was like pulling teeth trying to get the -g flag out.
I believe that this is a default behavior of autoconf scripts.
I remember having done some ugly hacks years ago to
25 matches
Mail list logo