Re: Lest we forget

From: Steve Long <answers_at_erols.com>
Date: 1997/01/30
Message-ID: <32F144BC.3120@erols.com>#1/1

tom,

gurry and corrigan in "Oracle Performance Tuning, 2nd Ed", pg 27, considering only RDBMS related issues, put it this way:

---

Causes of performance problems:

   Programs - 60.0% [referring to poor SQL and poor use of the optimizer]

   Design   - 20.0% [referring to data architecture]
   Database - 17.5%
   System   -  2.5%
--------------------------

Being of the software engineering type taking a total system perspective, I include program design in with database design, skewing the above numbers proportionately. My take on it is this:

    Design - 70.0% (referring to application and data architecture)     Programs - 20.0% (referring to overall coding environment unrelated to SQL)

    Database - 8.0%
    System - 2.0%

I stand by my assessment. You are welcome to disagree. We then just have to agree to disagree.

tdrudy_at_ix.netcom.com wrote:
>
> Steve,
> Respectfully, that depends on the OS and the hardware. 70%
> in NT, Solaris, HPUX, DGUX, Win95...??? 70% on a Sun Enterprise 6000
> with 1 Gig RAM or 14 Gigs RAM? On a SMP machine or single processor?
> With processors at 50Mhz or 167Mhz or Pentium Pros? RISC 6000 machines?
> RAID level 1 or 5 or no RAID? What kind of microcode has the OS been
> tweaked with, if any? Has it been tuned for a solely rdbms app (which
> can be done)? This is by no means a slam, but I'd like to hear from
> other Oracle dba's who are also Systems Administrators and Systems
> Engineers with assembler and OS experience.
>
> Tom
> Steve Long wrote:
> >
> > If I may add to this, 70% of all performance is in the data
> > architecture.
> >
> > -------------
> >
> > Roger Snowden wrote:
> > >
> > > Good points. Don't forget, however, that most of the performance issues
> > > for the typical transaction processing client/server system are application
> > > related. Like, runaway queries, full table scans, cartesian joins,
> > > non-selective indices, etc. Sometimes we get all caught up in the intrigue
> > > of OS and server tuning when the real problem is all the crap we are
> > > throwing at the engine to begin with.
> > >
> > > Just a thought.

--

Received on Thu Jan 30 1997 - 00:00:00 CST