Re: PostgreSQL slammed by PHP creator - Mailing list pgsql-general

From Arturo Pérez
Subject Re: PostgreSQL slammed by PHP creator
Date
Msg-id aperez-38C8C3.21143917092006@news.hub.org
Whole thread Raw
In response to PostgreSQL slammed by PHP creator  ("Arturo Perez" <aperez@hayesinc.com>)
List pgsql-general
In article <200609152029.00193.xzilla@users.sourceforge.net>,
 xzilla@users.sourceforge.net (Robert Treat) wrote:

> On Thursday 14 September 2006 12:19, Joshua D. Drake wrote:
> > Arturo Perez wrote:
> > > Hi all,
> > >
> > > Any response to this:
> > > http://www.internetnews.com/dev-news/article.php/3631831
> > >
>
> Well first, your subject line is very incendiary and not necessary... I

Well, sorry about that.  The article, as written, made it sound like he
said, "First, dump PostgreSQL to increase your performance then..."
>
> That said, in case anyone needs it, here is a benchmark showing postgresql
> scalability vs mysql.  interestingly enough it isnt really a database
> benchmark, whether this adds or detracts from it's creditability is up to
> you... http://tweakers.net/reviews/638/4

That's scalability, not speed.  Don't know that the PHP/MySQL crowd have
ever crowed about anything except speed.  For those not knowing the
difference, scalability is "we can handle 5000 concurrent transactions
with an average response time of 10ms" while speed is "we can make this
transaction complete in 10ms."  The difference being that something fast
may fall over if too many transactions occur within too small a window
(deadlocking, thread synchronization overhead, table/row locking, etc)
so that at 5000 concurrent transactions the speed may be reduced to
significantly (as shown in the tweakers.net pages).

> >
> > However, he does carry some umpf in certain circles. Perhaps we should
> > prove him wrong?
> >
>
> I'd like to see you do that... here are the slides from his recent talk
> showing  why he came up with the statements he made
> (http://talks.php.net/show/oscon06/1).  Please post the info
> when you get comparable performance running from PostgreSQL...

That's what I'd really like to know.  Honestly, how does one make pgSQL
go really really fast on a vanilla machine that one is likely to
encounter at a hosting facility?

As a pgSQL enthusiast, I want to see all the things needed to get pgSQL
to the 1500 tps level.  Unfortunately, we don't know anything about the
database requirements other than the one query shown in the source
examples.

I'll have to look at the system I recently converted from MySQL to
PostgreSQL and see what I can find...

-arturo

pgsql-general by date:

Previous
From: "Merlin Moncure"
Date:
Subject: postgresql rising
Next
From: "Samad, Alex"
Date:
Subject: Clustering