Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Mailing Lists -> Oracle-L -> RE: Database performance monitoring tool for developers.

RE: Database performance monitoring tool for developers.

From: Cary Millsap <cary.millsap_at_hotsos.com>
Date: Wed, 16 Apr 2003 13:09:14 -0800
Message-ID: <F001.0058350C.20030416130914@fatcity.com>


My $0.02...

Developers usually don't have access to a high-concurrency test environment in which an expensive monitoring tool would make that much difference anyway. The most powerful tools I've seen?

  1. Performance specifications - Functional specs contain a budgeted number of LIO operations that the code is allowed to consume. A good rule of thumb is 10 LIOs per (non-aggregated) result set row per table. For example, a 4-way join returning one row gets a budget of no more than 40 LIOs.
  2. Profiling (tkprof, autotrace, etc.) - EVERY piece of code gets traced and run through a profiler. If a piece of code breaks its specified LIO budget, then it's not approved for check-in.
  3. Execution plan analysis (explain plan) - EVERY piece of code has its execution plan checked by a performance analyst. Execution plans are generated with PRODUCTION db statistics, not test db statistics.
  4. The Wall of Shame - Write inefficient code, and your code goes up on the Wall for everybody to see. People whose names appear continually on the Wall of Shame are not selected as candidates for reproduction.

Who is the referee in all this? The performance analyst. Some performance analysts are DBAs, some are developers, some are architects, and some are none of the above. It doesn't matter who takes on the role, as long as it's someone competent and credible.

Cary Millsap
Hotsos Enterprises, Ltd.
http://www.hotsos.com

Upcoming events:

- Hotsos Clinic 101 in London, Reykjavik, Ottawa, Dallas, Denver, Sydney
- Visit www.hotsos.com for schedule details...
- IOUG-A Live 2003, Orlando, 10am Monday 28 April: "Oracle Operational
Timing Data"

-----Original Message-----
Faroult
Sent: Wednesday, April 16, 2003 3:24 PM
To: Multiple recipients of list ORACLE-L

"Grabowy, Chris" wrote:
>
> Yes, they are teaching you what to look for...
>
> > 1) Dictionary Cache Hits (ARGH!!!!!!!!!!!!!!!!!)
> > 2) Percentage of Current Processes (as in we have 30 processes
> > attached and 300 processes configured in the init.ora)
> > 3) Sessions Waiting for Lock
> > 4) Total Sort Rate (rate of sorts per minute) on disk and in memory)
>
> ;)
>
> And in my dealings with account execs, to close a deal, the hosting AE
> will make sure the client gets whatever reports they want...
>
> Can I go back into my corner now?
>
> -----Original Message-----
> Sent: Wednesday, April 16, 2003 11:29 AM
> To: Multiple recipients of list ORACLE-L
>
> it should be the hosting company's job to teach the clients what to
> really be looking at.
>
> --- "Grabowy, Chris" <cgrabowy_at_fcg.com> wrote:
> > (stumbling out of my lurker corner)
> >
> > Is it possible that some clients expect to see those cache hit ratio
> > reports? Arent there still quite a few Oracle sites that are still
> > hung up on RBO and cache hit ratios??
> >
> > -----Original Message-----
> > Sent: Wednesday, April 16, 2003 9:44 AM
> > To: Multiple recipients of list ORACLE-L
> >
> >
> > Ah but see, I didn't say 'DBA privs'. Most of the tools can be used by
> > an account that has select only on any of the v$ or dba_ views.
> >
> > I also said "database aware". Which means they understand what the
> > results mean.
> >
> > And then there are things like....
> >
> > the hosting company we use has reports on performance etc that
> > authorized users can look at. Reports are on CPU, Webservers, page
> > downloads and Oracle. Here's the sad part. There are 4 Oracle
> > reports:
> >
> > 1) Dictionary Cache Hits (ARGH!!!!!!!!!!!!!!!!!)
> > 2) Percentage of Current Processes (as in we have 30 processes
> > attached and 300 processes configured in the init.ora)
> > 3) Sessions Waiting for Lock
> > 4) Total Sort Rate (rate of sorts per minute) on disk and in memory)
> >
> > NONE of these are really useful, as they are reported on a daily
> > basis.
> >
> >
> > The first we know is really useless, and one of these days I'm going
> > to install a cron job to run Connor's BCHR script, just for giggles.
> >
> > The percentage of current processes makes no sense to me, as we run
> > with connection pooling.
> >
> > Sessions waiting for Lock on a daily basis? I suppose it could be
> > useful, if only to point me to an app to start tuning.
> >
> > And what really worries me is that the Sort Rate report is the "most
> > popular" (most viewed) of all of them. Not by me, I just learned about
> > these reports yesterday.
> >
> > I need to have a LONG talk with the hosting company.
> >
> >
> > --- April Wells <awells_at_csedge.com> wrote:
> > >
> > > I will go along with that.
> > > Whole heartedly in fact
> > > But with a disclaimer.
> > > They (the users with access to the tools) have to be willing to
> > learn
> > > the
> > > tool, and what the results MEAN. If they are willing to learn
> > > (really learn, not just hear part then think they know all there is
> > > to know about
> > > tuning code), I will give them access to the tools in a min.
> > >
> > > We have one developer who I trust implicitly. She wants to learn,
> > > tries to use good coding standards and tries to write the tightest
> > > code possible. I
> > > would give her DBA privs on the development instance in a
> > heartbeat,
> > > and any
> > > tool at my disposal to help her do her job.
> > >
> > > BUT...
> > >
> > > When you deal with parts of the user base (some of the developers)
> > > that think it is really cool to have a tool that has a really neat
> > > button that
> > > does everything for them, so they don't have to worry about the
> > > details of
> > > their code (SQL Navigator is our company's code generator of
> > > choice... not
> > > toad, that would be silly), then you have to deal with them having
> > > been
> > > 'educated' about tuning and what they should tell the DBAs to do to
> > > the
> > > system to tune it... and then they get an ORA-4030 error and decide
> > > that
> > > there needs to be a UNIX Kernel change because that will make their
> > > code
> > > run... you find out that a little bit of knowledge is a truly
> > > horrible
> > > thing.
> > >
> > >
> > > April
> > > -----Original Message-----
> > > To: Multiple recipients of list ORACLE-L
> > > Sent: 4/16/2003 5:58 AM
> > >
> > > If the developer is database-aware enough to want to use something
> > > like Spotlight, or any tool that does not degrade performance, I'll
> >
> > > give it
> > > to them if I can.
> > >
> > >
> > > --- Shamita & Chiran Ghosh <shamita-chiran_at_attbi.com> wrote:
> > > > Hello, How does the list feel about providing developers access
> > to
> > > > ad-hoc/real-time, problem solving diagnostics tool such as
> > > Spotlight,
> > > > Mamba
> > > > etc. to developers to provide first-line support. I'm not so
> > crazy
> > > > about the it; but was wondering on what the list's opinion would
> > be.
> >
> > > > Comments please.
> > > > Thanks.
> > > >
> > > >
> > > > Chiran Ghosh
> > > >
> > > > email: Shamita-Chiran_at_attbi.com

I have followed this thread with interest, and I fear that I don't share the usual enthusiasm for putting monitoring tools in the hands of developers - *some* developers, certainly, but not as a general rule. As Rachel pointed out, many monitoring tools, and sometimes purely for reasons of 'give the customer what s/he wants to get for his/her money', display mostly irrelevant values. Even if the values were relevant, would a young developer be able to tell what is abnormal, and what is just the normal consequence of crazy requirements, especially among a huge number of values, many of which are different symptoms of a same illness? Not so sure. I am fully in favour of educating developers, and I fear that very often throwing a tool to them may just be an excuse for not teaching them. I would be content enough if they were just able to feel that something is not as it ought to be.

Regards,

Stephane Faroult
Oriole Software

-- 
Please see the official ORACLE-L FAQ: http://www.orafaq.net
-- 
Author: Stephane Faroult
  INET: sfaroult_at_oriole.com

Fat City Network Services    -- 858-538-5051 http://www.fatcity.com
San Diego, California        -- Mailing list and web hosting services
---------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru_at_fatcity.com (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB ORACLE-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).

-- 
Please see the official ORACLE-L FAQ: http://www.orafaq.net
-- 
Author: Cary Millsap
  INET: cary.millsap_at_hotsos.com

Fat City Network Services    -- 858-538-5051 http://www.fatcity.com
San Diego, California        -- Mailing list and web hosting services
---------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru_at_fatcity.com (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB ORACLE-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).
Received on Wed Apr 16 2003 - 16:09:14 CDT

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US