Re: Evaluating client processes vs stored procedures

From: "codeWarrior" <gpatnude(at)hotmail(dot)com>
To: pgsql-general(at)postgresql(dot)org
Subject: Re: Evaluating client processes vs stored procedures
Date: 2006-04-12 21:51:36
Message-ID: e1jsml$7l9$1@news.hub.org
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-general

During your process loop -- when / where are the updates committed ? all at
the end ?

How may rows (approx) are you updating ?

FWIW:
I think you will probably find that it is NOT the SQL update that is your
bottleneck.... I am inclined to speculate that the performance issue is
related to the size of the loop or related to when / where you commit your
updates...

"Tim Hart" <tjhart(at)mac(dot)com> wrote in message
news:4684903(dot)1144871648075(dot)JavaMail(dot)tjhart(at)mac(dot)com(dot)(dot)(dot)
> List,
>
> I have a client process that is causing some performance issues with my
> app. On my current platform, the standard automated profiling tools are
> not available. Consequently, I can't easily determine where the bottleneck
> is in the process.
>
> I have an inkling that rewriting the client process to a stored procedure
> (implemented in C), will give at least a modest speed boost.
>
> What I'm trying to evaluate is the risk proposition of my effort. I really
> need a big bang for the buck, and I'm willing to risk a week of effort if
> the general consensus is that it's a good bet. I estimate a week to
> re-implement as a stored procedure, vs about 3 to manually profile what I
> need. Will I find out in 3 weeks what I already suspect? If I do profile
> it and fine-tune the slow points, would the newly architected client STILL
> run faster as a stored proc?
>
> So - any sage words of wisdom for those who've been down this path? I
> realize that w/o specifics no one can say for sure. General feelings from
> those with experience are welcome.
>
> Process specifics:
>
> Currently, the client process is run once a night. It's run on a separate
> machine than the database. Values in one table are updated by evaluating
> user defined functions against values in another table. Here's the
> pseudo-code:
>
> For each formula_row in formula_table:
> newValue = evaluate(formula_row.body)
> update formula_row set value = newValue where sid = formula_row.sid
>
>
> The evaluate function is a typical lexx/yacc parser. The formula body may
> refer to other formulas in the table, or to constant values in another
> table.
>
> If I do implement it as a stored proc, I'd likely add triggers so that the
> formulas would be re-evaluated once values in the constants table get
> updated.
>
> ---------------------------(end of broadcast)---------------------------
> TIP 1: if posting/reading through Usenet, please send an appropriate
> subscribe-nomail command to majordomo(at)postgresql(dot)org so that your
> message can get through to the mailing list cleanly
>

In response to

Browse pgsql-general by date

  From Date Subject
Next Message codeWarrior 2006-04-12 21:52:04 Re: "Hidden" field for each column
Previous Message Nick Johnson 2006-04-12 20:53:17 pg_restore --schema always returns an empty dump