Re: slow cursor

From: "Sriram Dandapani" <sdandapani(at)counterpane(dot)com>
To: "Benjamin Krajmalnik" <kraj(at)illumen(dot)com>, "Jim C(dot) Nasby" <jnasby(at)pervasive(dot)com>
Cc: <pgsql-admin(at)postgresql(dot)org>
Subject: Re: slow cursor
Date: 2006-04-21 04:36:42
Message-ID: 6992E470F12A444BB787B5C937B9D4DF0406AD8A@ca-mail1.cis.local
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-admin

Thanks Benjamin

The issue I have with my situation is tht the target table typically has
millions of records (not unusual to have tens of millions) and the
update is based on 80% of the columns. The update is a killer w.r.t
performance. I plan to create an index on all columns and test the
individual update.

Regards

Sriram

________________________________

From: Benjamin Krajmalnik [mailto:kraj(at)illumen(dot)com]
Sent: Thursday, April 20, 2006 9:10 PM
To: Sriram Dandapani; Jim C. Nasby
Cc: pgsql-admin(at)postgresql(dot)org
Subject: RE: [ADMIN] slow cursor

Siriam,

I have no clue what it is you are trying to do, but I have a similar
situation. I have a monitoring system which sends data to our
psotgresql database. The system calls a single stored procedure which
performs on-the-fly data aggregation for the particular test for which
the data is being sent. You may want to look at some numerical methods
books and see if you can do what I am doing. I keep a record with
computed values representing the data up to the previous transaction,
and recompute accordingly. This is done estremely quickly. Then, twice
a day, I run a stored procedure which runs some other aggregations based
on the raw data which was sent over. It was taking about 30 seconds
against a table with about 500K records, aggregating to a table with 2K
records. Since not all of the tests need this particular data
aggregation, an additional field was added to the test descriptor to
flag those tests which need to have the particular aggregation
(essentially, computing statistical process control data). After doing
this, our 30 seconds went down to about 2 seconds.

I do not have access to our aggregation dunction, but I am using one
cursor and 2 record types.

First, I perform a loop on a selection into the record structure for the
candidate tests. Within this loop, I create a cursor which fetches the
row data into another record structure. This record structure now holds
the data we need in order to update the table on which the primary loop
is running.

I do not know if this is the optimum way of doing this within
PostgreSQL, but the perofrmance appears to be fine. I have not goten
into the fine tuning for speed just yet.

Just an idea - hope it helps.

________________________________

From: pgsql-admin-owner(at)postgresql(dot)org on behalf of Sriram Dandapani
Sent: Thu 4/20/2006 9:45 PM
To: Jim C. Nasby
Cc: pgsql-admin(at)postgresql(dot)org
Subject: Re: [ADMIN] slow cursor

Yes..all of it is in one transaction as there is a window of record ids
that need to be processed in 1 transaction. Data inflow is very
voluminous appx 1 million every 15 minutes and the goal is to create
aggregate tables on the fly (the alternative is to use nightly
aggregates).

-----Original Message-----
From: Jim C. Nasby [mailto:jnasby(at)pervasive(dot)com]
Sent: Thursday, April 20, 2006 7:36 PM
To: Sriram Dandapani
Cc: pgsql-admin(at)postgresql(dot)org
Subject: Re: [ADMIN] slow cursor

On Mon, Apr 17, 2006 at 07:07:54AM -0700, Sriram Dandapani wrote:
> I have a cursor that fetches 150K rows and updates or inserts a table
> with 150K rows.
>
> It takes several minutes for the process to complete (about 15
minutes).
> The select by itself (without cursor) gets all rows in 15 seconds.
>
> Is there a way to optimize the cursor to fetch all records and speed
up
> the process. I still need to do the record by record processing

Not likely. Are you at least doing all this inside a transaction?
--
Jim C. Nasby, Sr. Engineering Consultant jnasby(at)pervasive(dot)com
Pervasive Software http://pervasive.com work: 512-231-6117
vcard: http://jim.nasby.net/pervasive.vcf cell: 512-569-9461

---------------------------(end of broadcast)---------------------------
TIP 4: Have you searched our list archives?

http://archives.postgresql.org

Browse pgsql-admin by date

  From Date Subject
Next Message Benjamin Krajmalnik 2006-04-21 04:43:23 Re: Howto: Using PITR recovery for standby replication
Previous Message Alvaro Herrera 2006-04-21 04:35:07 Re: Howto: Using PITR recovery for standby replication