Home » RDBMS Server » Performance Tuning » NOLOGGING in Oracle 11gr2 (Oracle 11gr2 Server 2008r2)
NOLOGGING in Oracle 11gr2 [message #568790] Tue, 16 October 2012 10:04 Go to next message
segan1415
Messages: 16
Registered: October 2012
Junior Member
I am building a database to store call quality statistics for VOIP networks. It is a very insert heavy application, and data reliability is of relatively minimal importance (in the sense that a few corrupt call records here and there doesn't matter the way corruption does in for example a banks database). Long term storage is also unimportant, most customers only wish to keep 3 months of data readily available in the database. Most do not even archive the older data.

To that end I am searching for every possible way to improve my insert performance and the internet has turned me onto the idea of NOLOGGING. These are the steps I have taken to reduce my IO consumed by the Redo and Undo logs.

1. I am inserting with the APPEND_VALUES hint.
2. I have disabled force logging at the database level
3. I have disabled force logging at the tablespace level
4. I have disabled logging on the relevant table and each of its indices

As best I can tell this is all I can do to minimize Redo/Undo, but based on my observations of the Disk portion of the WinServer2008 Performance Monitor, this has made little to no change in the amount of IO to my REDO and UNDO files. IO to the .dbf containing my table makes up less than 20% of the total disk IO for oracle.exe, the rest is the REDO and UNDO logs.

asktom.oracle.com/pls/asktom/f?p=100:11:0::::P11_QUESTION_ID:897564200346274711

The above article is a little over my head but I am able to extract from it that I will never entirely eliminate REDO/UNDO, which is fine, but I would think I could get it lower than it currently is.

So then, have I missed a step? Am I approaching this in entirely the wrong way? Other avenues to pursue? Any thoughts or advice would be appreciated.

Report message to a moderator

Re: NOLOGGING in Oracle 11gr2 [message #568791 is a reply to message #568790] Tue, 16 October 2012 10:22 Go to previous messageGo to next message
John Watson
Messages: 8960
Registered: January 2010
Location: Global Village
Senior Member
If performance is more important than reliability, you could set these parameters,
commit_wait=nowait;
commit_write=batch;
which will not eliminate the redo but will make it's generation faster.

But before doing anything like that, I would want to do more investigation. For example, the performance issue might be nothing to do with redo and undo: it could be (for example) concurrent access to the same blocks. Do you have AWR or statspack reports that you can post here?

--
Edit: Not commit_write=batch, but commit_logging=batch. Sorry, the parameter was changed between 10g and 11g.

[Updated on: Tue, 16 October 2012 10:27]

Report message to a moderator

Re: NOLOGGING in Oracle 11gr2 [message #568793 is a reply to message #568791] Tue, 16 October 2012 10:53 Go to previous messageGo to next message
segan1415
Messages: 16
Registered: October 2012
Junior Member
Here's an AWR report, in case you cannot read formatting, screenshot is attached.

Name SYSTEM_MOVING_WINDOW
ID 0
Type MOVING_WINDOW
Adaptive Thresholds Enabled? No
Window Size (Days) 8
Time Interval
Start Time 10/8/12 12:00 PM
End Time 10/16/12 11:00 AM
Start Snap ID 621
End Snap ID 795

Name Value Per Second Per Transaction
DB cpu (seconds) 0.00 0.00 0.00
DB time (seconds) 1,818,695.87 0.00 7.01
db block changes 71,467,189.00 0.00 275.39
execute count 6,090,428.00 0.00 23.47
global cache cr block receive time (seconds) 0.00 0.00 0.00
global cache cr blocks received 0.00 0.00 0.00
global cache current block receive time (seconds) 0.00 0.00 0.00
global cache current blocks received 0.00 0.00 0.00
global cache get time (seconds) 0.00 0.00 0.00
global cache gets 0.00 0.00 0.00
opened cursors cumulative 5,375,922.00 0.00 20.72
parse count (total) 2,495,161.00 0.00 9.61
parse time cpu (seconds) 221.72 0.00 0.00
parse time elapsed (seconds) 227.95 0.00 0.00
physical reads 239,314.00 0.00 0.92
physical writes 1,956,667.00 0.00 7.54
redo size (KB) 13,521,722.69 0.00 52.10
session cursor cache hits 2,870,743.00 0.00 11.06
session logical reads 114,941,958.00 0.00 442.91
sql execute cpu time (seconds) 0.00 0.00 0.00
sql execute elapsed time (seconds) 0.00 0.00 0.00
user calls 1,997,287.00 0.00 7.70
user commits 259,481.00 0.00 1.00
user rollbacks 34.00 0.00 0.00
workarea executions - multipass 0.00 0.00 0.00
workarea executions - onepass 1.00 0.00 0.00
workarea executions - optimal 767,909.00 0.00 2.96
  • Attachment: awr.png
    (Size: 49.60KB, Downloaded 1652 times)

Report message to a moderator

Re: NOLOGGING in Oracle 11gr2 [message #568794 is a reply to message #568793] Tue, 16 October 2012 10:58 Go to previous messageGo to next message
John Watson
Messages: 8960
Registered: January 2010
Location: Global Village
Senior Member
That is (a) illegible, please read How to use [code] tags and make your code easier to read and (b) not an AWR report. To generate an AWR report, run the script $ORACLE_HOME/rdbms/admin/awrrpt.sql

Report message to a moderator

Re: NOLOGGING in Oracle 11gr2 [message #568795 is a reply to message #568794] Tue, 16 October 2012 11:05 Go to previous messageGo to next message
BlackSwan
Messages: 26766
Registered: January 2009
Location: SoCal
Senior Member
Please read and follow the forum guidelines, to enable us to help you:

http://www.orafaq.com/forum/t/88153/0/

>Start Time 10/8/12 12:00 PM
>End Time 10/16/12 11:00 AM

any multi-day "snapshot" is multiple order of magnitude too long.

snapshot duration should be around 15 or 20 minutes at most!

Report message to a moderator

Re: NOLOGGING in Oracle 11gr2 [message #568796 is a reply to message #568795] Tue, 16 October 2012 11:21 Go to previous messageGo to next message
segan1415
Messages: 16
Registered: October 2012
Junior Member
Apologies on all accounts. I am working on creating a proper AWR report with awrrpt.sql in SQLDeveloper but am having difficulty at the moment. At first it complained it could not find awrrpti.sql in C:\Users\Administrators - I don't know why its looking there, but I eventually copied all the relevant scripts from my Oracle home to said directory in order to please it. At this point the script executes but eventually throws this error:

ORA-20200: Database/Instance /1 does not exist in DBA_HIST_DATABASE_INSTANCE
ORA-06512: at line 23


I am investigating this error and will get back to you once I have a proper report. Thanks for your help so far.

----
And for the record:

SELECT INSTANCE_NUMBER FROM v$instance; returns 1 so I have to believe 1 is correct.

----
Though on closer examination it appears that somehow my dbid is not being set correctly. At first I interpreted the '/1' to be a complaint regarding the Instance Number but now I realize it is a complaint about a null DBID.

[Updated on: Tue, 16 October 2012 11:42]

Report message to a moderator

Re: NOLOGGING in Oracle 11gr2 [message #568798 is a reply to message #568794] Tue, 16 October 2012 12:47 Go to previous messageGo to next message
segan1415
Messages: 16
Registered: October 2012
Junior Member
Well I'm sort of at wits end here. I believe I have successfully created a .html AWR report of a 15 minute window, through the Oracle Enterprise Manager web interface. However I don't know how to easily share an .html file with you, the forum does not allow them as attachments - I suppose I can put it in a dropbox if you are willing.

After that I continued to explore the error (ORA-20200) I'm getting in SQL developer. I am now skipping the execution of awrrpt.sql entirely and am instead hardcoding my variables into awrrpti.sql - in the case of dbid and instance number I have had to place them in awrinput.sql. However:


:dbid     :=  2907699466;


results in a

Numeric Overflow


error. Which is part of what I was seeing in SQL Developer (though I disregarded it at the time) and what I now believe is the reason my DBID is returning NULL instead of what I set it to in the prompt. So my DBID is out of range? That does not seem reasonable.



Heres the full output of @@awrrpti with a few embedded comments of my own:

Specify the Report Type
~~~~~~~~~~~~~~~~~~~~~~~
Would you like an HTML report, or a plain text report?
Enter 'html' for an HTML report, or 'text' for plain text
Defaults to 'html'
line 159: SQLPLUS Command Skipped: set heading off;
'TYPESPECIFIED:' REPORT_TYPE 
---------------- ----------- 
Type Specified:  text        
line 161: SQLPLUS Command Skipped: set heading on;
line 47: SQLPLUS Command Skipped: set heading on;
line 50: SQLPLUS Command Skipped: set trimspool on trimout on define "&" concat "." serveroutput on;
line 51: SQLPLUS Command Skipped: set underline on;


Instances in this Workload Repository schema
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
DBBID                                      INSTT_NUM              DBB_NAME  INSTT_NAME       HOST                                                             
------------------------------------------ ---------------------- --------- ---------------- ---------------------------------------------------------------- 
* 2907699466                               1                      GLDB      gldb             WIN-VTBNMB5VC5I                                                  

Using 6 for database Id         $$$$ When it prompted me I supplied 6 for both these
Using 6 for instance number     $$$$ However you can see in the next few lines I hardcode
                                $$$$ them to 2907699466 / 1, which is what I want. Then
                                $$$$ this triggers the Numeric Overflow and :dbid remains NULL

Error starting at line 91 in command:
begin
  :dbid      :=  2907699466;
  :inst_num  :=  1;
end;
Error report:
Numeric Overflow
Commit

Error starting at line 105 in command:
declare

  cursor cidnum is
     select 'X'
       from dba_hist_database_instance
      where instance_number = :inst_num
        and dbid            = :dbid;

  cursor csnapid is
     select to_char(max(end_interval_time),'dd/mm/yyyy')
       from dba_hist_snapshot
      where instance_number = :inst_num
        and dbid            = :dbid;

  vx     char(1);

begin

  -- Check Database Id/Instance Number is a valid pair
  open cidnum;
  fetch cidnum into vx;
  if cidnum%notfound then
    raise_application_error(-20200,
      'Database/Instance ' || :dbid || '/' || :inst_num ||
      ' does not exist in DBA_HIST_DATABASE_INSTANCE');
  end if;
  close cidnum;

  -- Check Snapshots exist for Database Id/Instance Number
  open csnapid;
  fetch csnapid into :max_snap_time;
  if csnapid%notfound then
    raise_application_error(-20200,
      'No snapshots exist for Database/Instance '||:dbid||'/'||:inst_num);
  end if;
  close csnapid;

end;
Error report:
ORA-20200: Database/Instance /1 does not exist in DBA_HIST_DATABASE_INSTANCE
ORA-06512: at line 23

[Updated on: Tue, 16 October 2012 12:55]

Report message to a moderator

Re: NOLOGGING in Oracle 11gr2 [message #568815 is a reply to message #568798] Tue, 16 October 2012 15:26 Go to previous messageGo to next message
John Watson
Messages: 8960
Registered: January 2010
Location: Global Village
Senior Member
I want to help you, but I can't! We need that AWR report. Log on to the server and set your environment (I see that it is Windows, so if you login as the Oracle owner the environment is probably setup for you through registry varaibles). Then connect to the database with SQL*Plus as a reasonably powerful user. Then run @?\rdbms\admin\awrrpt.sql and take the option to generate a text report. Choose two snapshots that cover the time span for a busy time, say one hour. And the report will be spooled out to your current directory. Post it here, and we'll have a look.

Report message to a moderator

Re: NOLOGGING in Oracle 11gr2 [message #568816 is a reply to message #568815] Tue, 16 October 2012 15:39 Go to previous messageGo to next message
segan1415
Messages: 16
Registered: October 2012
Junior Member
Ah thank you for the simpler instructions on how to generate such a report. It is attached.


----
In my own research I have found that Log File switching is occupying 19.12% of my dB time and occuring ~430 times per hour. I am going to increase my redo file sizes and see how performance is affected (currently I have 3 redo logs each of size = 51200KB). I will let you know what I see. Please let me know your thoughts on the report in any case, I'm sure there are many things I have missed.

[Updated on: Tue, 16 October 2012 16:24]

Report message to a moderator

Re: NOLOGGING in Oracle 11gr2 [message #568844 is a reply to message #568816] Wed, 17 October 2012 02:17 Go to previous messageGo to next message
Roachcoach
Messages: 1576
Registered: May 2010
Location: UK
Senior Member
For those who cannot download, contents in the spoiler section to save a wall of text Smile

Toggle Spoiler



At a quick glance, the database doesn't seem stressed - your log related events are high percentages because there's not a lot else going on (That horrible looking 19% is actually only a 62s of wall clock time in a 900s period). Have the clients raised a specific issue or is this a proactive tuning piece/early concern around scalability?

I suspect your plan to increase redo log size and add more groups will help reduce these values though.

Report message to a moderator

Re: NOLOGGING in Oracle 11gr2 [message #568850 is a reply to message #568816] Wed, 17 October 2012 02:39 Go to previous messageGo to next message
John Watson
Messages: 8960
Registered: January 2010
Location: Global Village
Senior Member
You are generating about 5M of redo per second, and log switching about every 10 seconds. This is because your online logfiules are sized at the old default of 50M, and is why you are losing time on those top wait events. If you replace your online logs with logfiles that are 500M, those wait events will disappear as your log switch frequency will drop to about 3 an hour.
Do that now! There is no downtime.
The next thing is more information: we need the AWR report for the problem sql. Run the script @?\rdbms\admin\sqlrpt.sql for the same time period, and give it the sql_id '7k1usds2tau95', which is your insert statement.

And lastly, the same reports after you have changed the redolog.

We can fix this.
--
Edit: typing error! Your online logfiles need to be 5000M, not 500M. Sorry about that.

[Updated on: Wed, 17 October 2012 03:09]

Report message to a moderator

Re: NOLOGGING in Oracle 11gr2 [message #568912 is a reply to message #568850] Wed, 17 October 2012 09:19 Go to previous messageGo to next message
segan1415
Messages: 16
Registered: October 2012
Junior Member
I think you meant sqltrpt.sql, not sqlrpt.sql? That is the only similar sql file I had so that is what I ran. It did not present me the option to pick a window of snapshots as did the other script, so this is of my database as it exists now with 3 5000M logfiles.

Toggle Spoiler


If you wish I can undo the changes I made to the log files and run it again.

Also I will shortly follow up with a new AWR report with my new log files. First I will produce a report with roughly the same traffic load as before. Then I will see if I can sustain more load - if I can I will produce a report for that as well.


Quote:
Have the clients raised a specific issue or is this a proactive tuning piece/early concern around scalability?


This is proactive tuning for future scalability.

Report message to a moderator

Re: NOLOGGING in Oracle 11gr2 [message #568913 is a reply to message #568912] Wed, 17 October 2012 09:22 Go to previous messageGo to next message
John Watson
Messages: 8960
Registered: January 2010
Location: Global Village
Senior Member
Sorry, the file is awrsqrpt.sql my aplogies

Report message to a moderator

Re: NOLOGGING in Oracle 11gr2 [message #568914 is a reply to message #568913] Wed, 17 October 2012 09:25 Go to previous messageGo to next message
segan1415
Messages: 16
Registered: October 2012
Junior Member
Ah ok. I will run that shortly then as well. I have begun the load transmission to generate the next batch of AWR and I want that to run uninterrupted as before. My Windows Performance Monitor looks much better already though, I think we are making great strides Razz

Report message to a moderator

Re: NOLOGGING in Oracle 11gr2 [message #568915 is a reply to message #568913] Wed, 17 October 2012 10:00 Go to previous messageGo to next message
segan1415
Messages: 16
Registered: October 2012
Junior Member
It seems from the new AWRRPT, I am now swapping log files every fifteen minutes (since there is only 1 switch, I ran a second awrrpt from 801-803 and there was 2 switches - confirming this)

AWRRPT 801-802 (50M REDO LOG FILES)
Toggle Spoiler



AWRRPT 887-888 (5000M REDO LOG FILES)
Toggle Spoiler


AWRSQLRPT 801-802 (50M REDO LOG FILES)
Toggle Spoiler



AWRSQLRPT 887-888 (5000M REDO LOG FILES)
Toggle Spoiler

[Updated on: Wed, 17 October 2012 10:08]

Report message to a moderator

Re: NOLOGGING in Oracle 11gr2 [message #568917 is a reply to message #568915] Wed, 17 October 2012 10:17 Go to previous messageGo to next message
John Watson
Messages: 8960
Registered: January 2010
Location: Global Village
Senior Member
As promised, we've knocked out those wait events. And the proportion of DBTime that is actually working has gone from 49.8% up to 64.5%. That is good, but the number of executes per second has not significantly changed, so as RoachCoach said, perhaps there isn't actually enough work to do. Are you certain that the bottleneck is not the application server? It looks to me as though the work is being done more efficiently: could it be that the DB is in fact processing data as fast as you can provide it?
Note that you are not in fact getting a direct load, you are getting a conventional insert. You need to move your hint, it must be between SELECT and INTO.

(ps - brilliant use of tags, you will make a lot of friends here if you keep it up)
--
Edit: another dumb typo, SELECT above shoud read INSERT. But you all know that, of course. jw.

[Updated on: Wed, 17 October 2012 10:35]

Report message to a moderator

Re: NOLOGGING in Oracle 11gr2 [message #568918 is a reply to message #568917] Wed, 17 October 2012 10:23 Go to previous messageGo to next message
John Watson
Messages: 8960
Registered: January 2010
Location: Global Village
Senior Member
And one more point: the sequence. How many values is it caching?
select cache_size,sequence_name from dba_sequences where sequence_name='GLS_PROT_TRAFFIC_SUM_VOIP';

if it is the default of 20, change it. Perhaps:
alter sequence GLS_PROT_TRAFFIC_SUM_VOIP cache 10000;

and that may help a lot.

Report message to a moderator

Re: NOLOGGING in Oracle 11gr2 [message #568920 is a reply to message #568918] Wed, 17 October 2012 10:38 Go to previous messageGo to next message
segan1415
Messages: 16
Registered: October 2012
Junior Member
Glad you like my tags Smile, and thanks for the pointer on my hint I will correct that. Yes the wait events have diminished dramatically which is fantastic, and performance has improved somewhat as a result. And you are right my cache_size is 20, I will up it accordingly and test again.

Regarding whether I'm sure where the bottleneck is, no I am definitely not sure. My basic system architecture is I have a number of probes distributed through a VOIP network (this is the real world architecture at least, in the lab I just have a single CLI application I use to generate steady, regular loads), and a central machine running both the Oracle RDBMS and an ODBC Loader application. The probes connect to the ODBC loader via TCP, and the loader connects to the database.

I know based on wireshark observation that the load generator is doing its job, and I am running this in a lab LAN at the moment with sub millisecond latency. So I feel safe in ruling that out as a bottleneck. The ODBC loader application accepts incoming records from the probes or load generator, caches them, and array inserts them into the database. By no means am I convinced that the ODBC loader is not the bottleneck, frankly I sort of have a hunch that it is. But my first job before I go digging into that code is to prove that Oracle itself is healthy.

Report message to a moderator

Re: NOLOGGING in Oracle 11gr2 [message #568921 is a reply to message #568920] Wed, 17 October 2012 10:53 Go to previous messageGo to next message
segan1415
Messages: 16
Registered: October 2012
Junior Member
Are you alarmed by the extraordinary number of Buffer Gets?

Plan Statistics                             DB/Inst: GLDB/gldb  Snaps: 887-888
-> % Total DB Time is the Elapsed Time of the SQL statement divided
   into the Total Database Time multiplied by 100

Stat Name                                Statement   Per Execution % Snap
---------------------------------------- ---------- -------------- -------
Elapsed Time (ms)                           261,736           10.8    90.1
CPU Time (ms)                               164,924            6.8    88.1
Executions                                   24,205            N/A     N/A
Buffer Gets                              3.2983E+07        1,362.7    98.1
Disk Reads                                        4            0.0     2.6
Parse Calls                                       1            0.0     0.0
Rows                                      2,415,000           99.8     N/A

Report message to a moderator

Re: NOLOGGING in Oracle 11gr2 [message #568923 is a reply to message #568921] Wed, 17 October 2012 10:59 Go to previous messageGo to next message
BlackSwan
Messages: 26766
Registered: January 2009
Location: SoCal
Senior Member
>Are you alarmed by the extraordinary number of Buffer Gets?

what is the alternative?
Would you prefer to do Disk Read instead?
From where else should the desired data blocks be obtained?

Report message to a moderator

Re: NOLOGGING in Oracle 11gr2 [message #568924 is a reply to message #568923] Wed, 17 October 2012 11:01 Go to previous messageGo to next message
segan1415
Messages: 16
Registered: October 2012
Junior Member
Well I don't completely understand most of what I'm looking at, I'm learning a lot of this on the fly. But if the alternative is disk reads then point taken.

Report message to a moderator

Re: NOLOGGING in Oracle 11gr2 [message #568925 is a reply to message #568924] Wed, 17 October 2012 11:13 Go to previous messageGo to next message
John Watson
Messages: 8960
Registered: January 2010
Location: Global Village
Senior Member
wrt the buffer gets, are there many indexes on that table? If you can lose any, that would reduce the buffer gets and therefore CPU per insert.
So far, you haven't really attacked the actual run time, only the wait events: CPU time per execute has dropped from 7.0 to 6.8, elapsed time from 13.8 to 10.8. Tuning the sequence (and given that it is millions of rows, better cache at least 100000) and dropping an index might help.
Any chance that you can test these things one at a time, also test the direct load individually?

Report message to a moderator

Re: NOLOGGING in Oracle 11gr2 [message #568926 is a reply to message #568925] Wed, 17 October 2012 11:32 Go to previous messageGo to next message
segan1415
Messages: 16
Registered: October 2012
Junior Member
Yes I have free reign over the dB, so I will test each of the following variables individually:

*direct load
*sequence cache size
*index count (currently there is the primary key and three additional indexes on the table)

This will take some time, but I'll post the results as soon as I have them.

And thanks for all the help so far, it is much appreciated.

[Updated on: Wed, 17 October 2012 14:24]

Report message to a moderator

Re: NOLOGGING in Oracle 11gr2 [message #568941 is a reply to message #568926] Wed, 17 October 2012 15:15 Go to previous messageGo to next message
segan1415
Messages: 16
Registered: October 2012
Junior Member
I have changed sequence cache to 100,000 per your recommendation. This change chops about .4ms from my elapsed time per execution. The first two reports (907-908) were run at the same load volume as all my previous reports, then I decided to run my load generator at a higher volume - a volume I knew the Loader/dB would not be able to handle without the loader overflowing and losing data - report 908-909. They look about the same to my eyes, which I think proves that the loader is the issue, and not the database. Not to say that the dB cannot be further tuned, simply that it is not the bottleneck. Is this a logical conclusion?

AWRRPT 907-908 (sustainable call volume)
Toggle Spoiler


AWRSQLRPT 907-908 (sustainable call volume)
Toggle Spoiler


AWRRPT 908-909 (unsustainable call volume)
Toggle Spoiler


AWRSQLRPT 908-909 (unsustainable call volume)
Toggle Spoiler

Report message to a moderator

Re: NOLOGGING in Oracle 11gr2 [message #568943 is a reply to message #568925] Wed, 17 October 2012 16:58 Go to previous messageGo to next message
segan1415
Messages: 16
Registered: October 2012
Junior Member
Here is reports for DIRECT PATH insert. My observations:
*REDO size reduced by two orders of magnitude, not surprising
*Essentially everything in the Load Profile is reduced by an order of magnitude or more on a per transaction
*On a per second basis much of the Load Profile is actually increased however.
*My CPU is spending less time waiting overall
*In the AWRSQLREPORT, by time per execution is cut by 3.9ms, Buffer gets are cut by a third and Disk Reads disappear entirely

So at first glance the numbers look good to me. But the end result is the same bottleneck as before. Further confirming the idea that Oracle is now essentially OK and I need to examine my loader and perhaps ditch ODBC entirely in favor of something else.

AWRRPT 913-914 (DIRECT PATH INSERT)
Toggle Spoiler


AWRSQRPT 913-914 (DIRECT PATH INSERT)
Toggle Spoiler

Report message to a moderator

Re: NOLOGGING in Oracle 11gr2 [message #568957 is a reply to message #568943] Thu, 18 October 2012 01:44 Go to previous messageGo to next message
John Watson
Messages: 8960
Registered: January 2010
Location: Global Village
Senior Member
An excellent result (though you would need to think about concurrency issues with direct load, partitioning might be necessary). At the statement level the buffer gets per execution is halved, and elapsed time drops close to CPU time. At the instance level, the DB Time required to execute the same workload has come down to 181s from originally 324s. And CPU time is up to 86.2% of the DB Time. I very much doubt that this can be tuned further (unless you were to drop some indexes). To put it in practical terms, you could run this application on a single core no problem. So you must be correct to say that the bottleneck is further back. I certainly can't advise on ODBC or writing OCI code, but there are people here who can. Separate topic!

Thanks for bearing with me, even though the DB was probably never a problem. It was an enjoyable exercise.

Report message to a moderator

Re: NOLOGGING in Oracle 11gr2 [message #569004 is a reply to message #568957] Thu, 18 October 2012 09:18 Go to previous message
segan1415
Messages: 16
Registered: October 2012
Junior Member
Yes I will explore ODBC and OCI in a new topic. Thanks very much for the help John and everyone else who commented, this was quite productive for me and I appreciate it very much.

Report message to a moderator

Previous Topic: Improve performance postgre
Next Topic: incremental stats not happening
Goto Forum:
  

[ Syndicate this forum (XML) ] [ RSS ]

Current Time: Thu Nov 21 16:51:49 CST 2024
.:: Forum Home :: Blogger Home :: Wiki Home :: Contact :: Privacy ::.