Re: Shared Pool info - V$sqlarea.

2001-06-06 Thread Mogens Nørgaard

He also gave the presentation at the IOUG-A in Orlando recently, so it
should be available from their CD's/archives or whatever. It's a neat
presentation. Way too many people and monitoring products want you do keep
increasing db_block_buffers until your BCHR is about 99.9% or so. It's
pure nonsense, of course.

Once I worked on a large SAP site, where the SAP Early Watch kept saying
the problems in the system were due to the BCHR being too low. That went
on for a few months while performance became progressivly worse. When the
customer passed the 1G mark (for the buffer cache alone) they called
me/Premium Services and the wait interface showed all the wait time to be
with the free buffer waits event, meaning the DBWR couldn't keep up (no
async IO, therefor making the buffer cache bigger made the problem bigger,
etc.). We set up multiple database writers to simulate async IO (long
story) and it all worked nicely...

This is not to say anything bad about Early Watch, but to say something
bad about the ratio-based/checklist tuning method :-))).

A. Bardeen wrote:

 Raj,

 One statement caught my eye:
   The buffer cache hit ratio is 99%

 My bet is you've got some very inefficent SQL.  As is
 typically the case, application tuning has the biggest
 effect on performance.

 I'd suggest checking out Cary Millsap's excellent
 paper Why 99% Database Buffer Cache Hit Ratio is NOT
 Ok (available with free registration):
   http://www.hotsos.com/catalog/mtime.html

 HTH,

 -- Anita

 --- Raj Gopalan [EMAIL PROTECTED]
 wrote:
  Thanks Chris.
 
  The problem I am facing is 100% CPU usage and memory
  paging out at times. I
  tought the starting point is v$sqlarea. Purchasing
  of Precise SQL or SQL
  Vision can not happen here immediately. The RAM size
  is 512MB and SGA is
  210MB. The buffer cache hit ratio is 99%.
 
  I was just wondering is there any way to find out
  the cause of this problem?
 
  TIA
 
  Cheers
 
  Raj
 
  -Original Message-
  Sent: 01 June 2001 19:51
  To: Multiple recipients of list ORACLE-L
 
 
  Oracle has no guarentee how long statistics and
  plans will be available for
  statements, depending on the activity of the
  database they may be there for
  2 seconds they may be there for 2 weeks.
 
  Products like Precise SQL and SQL Vision Lab help in
  that manor where they
  capture transactions and activity continuously in a
  non-intrusive manor.
  This is the only guarenteed way to get 99.999% of
  the transactions.
 
 
  Walking on water and developing software from a
  specification are easy if
  both are frozen.
 
  Christopher R. Spence
  Oracle DBA
  Fuelspot
 
 
 
  -Original Message-
  Sent: Friday, June 01, 2001 12:16 PM
  To: Multiple recipients of list ORACLE-L
 
 
  DBAs
 
  The statistics in v$sqlarea is getting flushed very
  often.
 
  In the morning I found the a query which has more
  than 10,000 disk reads as
  the top one. But now the top most query with disk
  reads has not more than
  100 disk reads. I have not bouncd the DB or flushed
  the shared pool. The
  overall library cache hit ratio is 97%.where us
  SQLAREA hit ratio is 60%.
 
  Any Idea what could be reason?
 
  TIA,
 
  Raj
  --
  Please see the official ORACLE-L FAQ:
  http://www.orafaq.com
  --
  Author: Raj Gopalan
INET: [EMAIL PROTECTED]
 
  Fat City Network Services-- (858) 538-5051  FAX:
  (858) 538-5051
  San Diego, California-- Public Internet
  access / Mailing Lists
 
 
  To REMOVE yourself from this mailing list, send an
  E-Mail message
  to: [EMAIL PROTECTED] (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.com
  --
  Author: Christopher Spence
INET: [EMAIL PROTECTED]
 
  Fat City Network Services-- (858) 538-5051  FAX:
  (858) 538-5051
  San Diego, California-- Public Internet
  access / Mailing Lists
 
 
  To REMOVE yourself from this mailing list, send an
  E-Mail message
  to: [EMAIL PROTECTED] (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.com
  --
  Author: Raj Gopalan
INET: [EMAIL PROTECTED]
 
  Fat City Network Services-- (858) 538-5051  FAX:
  (858) 538-5051
  San Diego, California-- Public Internet
  access / Mailing Lists
 
 
  To REMOVE yourself from this mailing list, send an
  E-Mail 

RE: Shared Pool info - V$sqlarea.

2001-06-05 Thread Raj Gopalan

Thanks Chris.

The problem I am facing is 100% CPU usage and memory paging out at times. I
tought the starting point is v$sqlarea. Purchasing of Precise SQL or SQL
Vision can not happen here immediately. The RAM size is 512MB and SGA is
210MB. The buffer cache hit ratio is 99%. 

I was just wondering is there any way to find out the cause of this problem?

TIA

Cheers

Raj

-Original Message-
Sent: 01 June 2001 19:51
To: Multiple recipients of list ORACLE-L


Oracle has no guarentee how long statistics and plans will be available for
statements, depending on the activity of the database they may be there for
2 seconds they may be there for 2 weeks.

Products like Precise SQL and SQL Vision Lab help in that manor where they
capture transactions and activity continuously in a non-intrusive manor.
This is the only guarenteed way to get 99.999% of the transactions.


Walking on water and developing software from a specification are easy if
both are frozen.

Christopher R. Spence
Oracle DBA
Fuelspot 



-Original Message-
Sent: Friday, June 01, 2001 12:16 PM
To: Multiple recipients of list ORACLE-L


DBAs

The statistics in v$sqlarea is getting flushed very often. 

In the morning I found the a query which has more than 10,000 disk reads as
the top one. But now the top most query with disk reads has not more than
100 disk reads. I have not bouncd the DB or flushed the shared pool. The
overall library cache hit ratio is 97%.where us SQLAREA hit ratio is 60%.

Any Idea what could be reason?

TIA,

Raj
-- 
Please see the official ORACLE-L FAQ: http://www.orafaq.com
-- 
Author: Raj Gopalan
  INET: [EMAIL PROTECTED]

Fat City Network Services-- (858) 538-5051  FAX: (858) 538-5051
San Diego, California-- Public Internet access / Mailing Lists

To REMOVE yourself from this mailing list, send an E-Mail message
to: [EMAIL PROTECTED] (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.com
-- 
Author: Christopher Spence
  INET: [EMAIL PROTECTED]

Fat City Network Services-- (858) 538-5051  FAX: (858) 538-5051
San Diego, California-- Public Internet access / Mailing Lists

To REMOVE yourself from this mailing list, send an E-Mail message
to: [EMAIL PROTECTED] (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.com
-- 
Author: Raj Gopalan
  INET: [EMAIL PROTECTED]

Fat City Network Services-- (858) 538-5051  FAX: (858) 538-5051
San Diego, California-- Public Internet access / Mailing Lists

To REMOVE yourself from this mailing list, send an E-Mail message
to: [EMAIL PROTECTED] (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).



Re: Shared Pool info - V$sqlarea.

2001-06-05 Thread Jared Still


Did you say 99% hit ratio?

Contrary to a once popular opinion, that is 
not really a good thing.

It often means there's a cartesian product
in a join, caused by not fully specifying
the join key. At least that's the reason
I've usually seen for it.

Keeps disk access down though. :)

Jared

On Tuesday 05 June 2001 05:25, Raj Gopalan wrote:
 Thanks Chris.

 The problem I am facing is 100% CPU usage and memory paging out at times. I
 tought the starting point is v$sqlarea. Purchasing of Precise SQL or SQL
 Vision can not happen here immediately. The RAM size is 512MB and SGA is
 210MB. The buffer cache hit ratio is 99%.

 I was just wondering is there any way to find out the cause of this
 problem?

 TIA

 Cheers

 Raj

 -Original Message-
 Sent: 01 June 2001 19:51
 To: Multiple recipients of list ORACLE-L


 Oracle has no guarentee how long statistics and plans will be available for
 statements, depending on the activity of the database they may be there for
 2 seconds they may be there for 2 weeks.

 Products like Precise SQL and SQL Vision Lab help in that manor where they
 capture transactions and activity continuously in a non-intrusive manor.
 This is the only guarenteed way to get 99.999% of the transactions.


 Walking on water and developing software from a specification are easy if
 both are frozen.

 Christopher R. Spence
 Oracle DBA
 Fuelspot



 -Original Message-
 Sent: Friday, June 01, 2001 12:16 PM
 To: Multiple recipients of list ORACLE-L


 DBAs

 The statistics in v$sqlarea is getting flushed very often.

 In the morning I found the a query which has more than 10,000 disk reads as
 the top one. But now the top most query with disk reads has not more than
 100 disk reads. I have not bouncd the DB or flushed the shared pool. The
 overall library cache hit ratio is 97%.where us SQLAREA hit ratio is 60%.

 Any Idea what could be reason?

 TIA,

 Raj
-- 
Please see the official ORACLE-L FAQ: http://www.orafaq.com
-- 
Author: Jared Still
  INET: [EMAIL PROTECTED]

Fat City Network Services-- (858) 538-5051  FAX: (858) 538-5051
San Diego, California-- Public Internet access / Mailing Lists

To REMOVE yourself from this mailing list, send an E-Mail message
to: [EMAIL PROTECTED] (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).



Re: Shared Pool info - V$sqlarea.

2001-06-05 Thread George Schlossnagle

Particularly, this can be caused by a huge amount of logical IO 
(proportionately dwarfing the amount of disk io and resulting in a 
falsely 'positive' hit ratio).  Take a look at your sqlarea again and 
look at buffer_gets/executions - that may kick up some queries doing 
more logical io than you want.

George
On Tuesday, June 5, 2001, at 10:40 PM, Jared Still wrote:


 Did you say 99% hit ratio?

 Contrary to a once popular opinion, that is
 not really a good thing.

 It often means there's a cartesian product
 in a join, caused by not fully specifying
 the join key. At least that's the reason
 I've usually seen for it.

 Keeps disk access down though. :)

 Jared

 On Tuesday 05 June 2001 05:25, Raj Gopalan wrote:
 Thanks Chris.

 The problem I am facing is 100% CPU usage and memory paging out at 
 times. I
 tought the starting point is v$sqlarea. Purchasing of Precise SQL or 
 SQL
 Vision can not happen here immediately. The RAM size is 512MB and SGA 
 is
 210MB. The buffer cache hit ratio is 99%.

 I was just wondering is there any way to find out the cause of this
 problem?

 TIA

 Cheers

 Raj

 -Original Message-
 Sent: 01 June 2001 19:51
 To: Multiple recipients of list ORACLE-L


 Oracle has no guarentee how long statistics and plans will be 
 available for
 statements, depending on the activity of the database they may be 
 there for
 2 seconds they may be there for 2 weeks.

 Products like Precise SQL and SQL Vision Lab help in that manor where 
 they
 capture transactions and activity continuously in a non-intrusive 
 manor.
 This is the only guarenteed way to get 99.999% of the transactions.


 Walking on water and developing software from a specification are 
 easy if
 both are frozen.

 Christopher R. Spence
 Oracle DBA
 Fuelspot



 -Original Message-
 Sent: Friday, June 01, 2001 12:16 PM
 To: Multiple recipients of list ORACLE-L


 DBAs

 The statistics in v$sqlarea is getting flushed very often.

 In the morning I found the a query which has more than 10,000 disk 
 reads as
 the top one. But now the top most query with disk reads has not more 
 than
 100 disk reads. I have not bouncd the DB or flushed the shared pool. 
 The
 overall library cache hit ratio is 97%.where us SQLAREA hit ratio is 
 60%.

 Any Idea what could be reason?

 TIA,

 Raj
 --
 Please see the official ORACLE-L FAQ: http://www.orafaq.com
 --
 Author: Jared Still
   INET: [EMAIL PROTECTED]

 Fat City Network Services-- (858) 538-5051  FAX: (858) 538-5051
 San Diego, California-- Public Internet access / Mailing Lists
 
 To REMOVE yourself from this mailing list, send an E-Mail message
 to: [EMAIL PROTECTED] (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.com
-- 
Author: George Schlossnagle
  INET: [EMAIL PROTECTED]

Fat City Network Services-- (858) 538-5051  FAX: (858) 538-5051
San Diego, California-- Public Internet access / Mailing Lists

To REMOVE yourself from this mailing list, send an E-Mail message
to: [EMAIL PROTECTED] (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).



Shared Pool info - V$sqlarea.

2001-06-01 Thread Raj Gopalan

DBAs

The statistics in v$sqlarea is getting flushed very often. 

In the morning I found the a query which has more than 10,000 disk reads as
the top one. But now the top most query with disk reads has not more than
100 disk reads. I have not bouncd the DB or flushed the shared pool. The
overall library cache hit ratio is 97%.where us SQLAREA hit ratio is 60%.

Any Idea what could be reason?

TIA,

Raj
-- 
Please see the official ORACLE-L FAQ: http://www.orafaq.com
-- 
Author: Raj Gopalan
  INET: [EMAIL PROTECTED]

Fat City Network Services-- (858) 538-5051  FAX: (858) 538-5051
San Diego, California-- Public Internet access / Mailing Lists

To REMOVE yourself from this mailing list, send an E-Mail message
to: [EMAIL PROTECTED] (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).