Sounds like an issue I have experienced in Oracle as well. If you can
you might want consider breaking out your database into oltp (on line
transaction processing) and data warehouse db. You run you any reports
you can nightly into a set of warehouse tables and save your daytime
cpus for incoming info and special real-time (hottest commodity of the
day) reports that you have tuned the best you can. Anything you can
calculate in advance that won't change over time, should be saved in the
warehouse tables, so you don't waste cpus, re-working data in real time.
Pre-running your reports won't speed them up but your users won't be
waiting for a report to calculate while they are looking at the screen. 

-----Original Message-----
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On Behalf Of Bill
Sent: Tuesday, June 22, 2004 11:31 AM
To: Josh Berkus
Cc: [EMAIL PROTECTED]
Subject: Re: [PERFORM] postgresql and openmosix migration

Ok, so maybe someone on this group will have a better idea.  We have a
database of financial information, and this has literally millions of
entries.  I have installed indicies, but for the rather computationally
demanding processes we like to use, like a select query to find the
commodity with the highest monthly or annual returns, the computer
generally
runs unacceptably slow.  So, other than clustring, how could I achieve a
speed increase in these complex queries?  Is this better in mysql or
postgresql?

Thanks.
----- Original Message ----- 
From: "Josh Berkus" <[EMAIL PROTECTED]>
To: "Bill" <[EMAIL PROTECTED]>; <[EMAIL PROTECTED]>
Sent: Tuesday, June 22, 2004 11:31 AM
Subject: Re: [PERFORM] postgresql and openmosix migration


> Bill,
>
> >   Any ideas of how I can cluster my database (around 800 GB
> > in size so even partial replication is not really practical)?
>
> Um, raise $150,000 to pay for a clustering implementation?
>
> Various techniques of "shared memory clustering" have been tried with
> PostgreSQL, and none work.    Neither does LinuxLabs "ClusGres", which
is
> based on similar principles -- unfortunately. (at least, LL repeatedly
> postponed the demo they said they'd give me.  I've yet to see anything
> working ...)
>
> Frankly, we're waiting for a well-funded corporation to jump in and
decide
> they want PostgreSQL clustering.   Database server clustering is a
"big
> ticket item" requiring roughly 1,000 hours of programming and
> troubleshooting.   As such, you're not likely to see it come out of
the
OSS
> community unaided.
>
> Oh, and FYI, MySQL's "clustering" doesn't work either.  It requires
your
> entire database to fit into available RAM ....
>
> -- 
> Josh Berkus
> Aglio Database Solutions
> San Francisco
>



---------------------------(end of broadcast)---------------------------
TIP 7: don't forget to increase your free space map settings

---------------------------(end of broadcast)---------------------------
TIP 6: Have you searched our list archives?

               http://archives.postgresql.org

Reply via email to