Re: Prefetch - Mailing list pgsql-performance

From Mischa Sandberg
Subject Re: Prefetch
Date
Msg-id 1115838416.428257d0e5bc3@webmail.telus.net
Whole thread Raw
In response to Re: Prefetch  (Christopher Kings-Lynne <chriskl@familyhealth.com.au>)
List pgsql-performance
Quoting Christopher Kings-Lynne <chriskl@familyhealth.com.au>:

> > Another trick you can use with large data sets like this when you
> want
> > results
> > back in seconds is to have regularly updated tables that aggregate
> the data
> > along each column normally aggregated against the main data set.
>
> > Maybe some bright person will prove me wrong by posting some
> working
> > information about how to get these apparently absent features
> working.
>
> Most people just use simple triggers to maintain aggregate summary
> tables...

Don't know if this is more appropriate to bizgres, but:
What the first poster is talking about is what OLAP cubes do.

For big aggregating systems (OLAP), triggers perform poorly,
compared to messy hand-rolled code. You may have dozens
of aggregates at various levels. Consider the effect of having
each detail row cascade into twenty updates.

It's particularly silly-looking when data is coming in as
batches of thousands of rows in a single insert, e.g.

   COPY temp_table FROM STDIN;
   UPDATE fact_table ... FROM ... temp_table
   INSERT INTO fact_table ...FROM...temp_table

   (the above pair of operations is so common,
    Oracle added its "MERGE" operator for it).

Hence my recent post (request) for using RULES to aggregate
--- given no luck with triggers "FOR EACH STATEMENT".


pgsql-performance by date:

Previous
From: Tom Lane
Date:
Subject: Re: Bad plan after vacuum analyze
Next
From: Guillaume Smet
Date:
Subject: Re: Bad plan after vacuum analyze