[oclug]PostgreSQL Hardware
Brad Barnett
bb at L8R.net
Mon Oct 28 09:06:55 EST 2002
On Mon, 28 Oct 2002 07:34:04 -0500
Milan Budimirovic <milan.budimirovic at sympatico.ca> wrote:
> Phil Tanguay wrote:
>
> > Well, first off, I don't really want to particpate in the "he said,
> > (s)he said" you guys are playing with, but I'll throw this in to the
> > equation :
> >
> > It seems to me that most often, when I see a web site "slashdotted",
> > the pages have lovely "mySql error: could not open connection" or
> > somesuch. So I tend to agree with Brad on this count, usually what
> > seems to fall over first is the DB server.
>
> That means absolutely nothing. MySQL has an environmental variable that
> sets the maximum number of connections, and the default value is very
> low (I don't recall what it is offhand, but it's generally an order of
> magnitude less than the server can handle -- like 100 vs. 1000+). When
> you see an error of that type you are invariably because you are butting
> your head up against an environmental setting, rather than the limits of
> the database server itself.
>
>
> > With that being said, the biggest issue here is that we don't have
> > enough data to be able to make an intelligent comments : the way to
> > design a system(and by system, I mean the combination of both the
> > hardware and software) is by designing the application, and THEN
> > picking the hardware and applying the most money to the components
> > that need it most. In my humble opinion, picking out the hardware
> > first is a rookie mistake. It unfortunately happens quite often,
> > because buying hardware and spending money is "fun" : analyzing
> > databases schemas and data flow, much less.
> >
> > But in this situation, with our lack of knowledge of the problem
> > domain, we're left to make blanket statements that are at best,
> > sometimes true, sometimes not. Much like I did in the first paragraph
> > of this post. ;>
> >
>
> It was my understanding that the application was aready built and that
> the database was being imported from MS SQL Server, so there shouldn't
> have been any major surprises here.
True, but don't blame Phil for no knowing. The only time he could have
read about that fact was about 3 weeks ago, in another thread. That's a
long time ago, in number of messages, on a list like this. I didn't
notice it (and I'm sure many others didn't). Good memory Milan ;)
Anyhow, one thing that should be mentioned in this venue is that since it
is coming from an MS SQL server, a retooling of the data and queries may
be in order. Every SQL server has its shining star. Postgres, for
example, will be quicker at certain types of queries than MSSQL is, and
less efficient with others.
The less complex the query, the less work needs to be done, but even in
that case, choose your column and table types wisely. Simply using the
same types from the MS SQL server may not be a good idea, in terms of fine
tuning performance.
More information about the OCLUG
mailing list