.comment-link {margin-left:.6em;}

Oracle Sponge -- Now Moved To Wordpress

Please use http://oraclesponge.wordpress.com

Monday, April 25, 2005

Thread about block sizes

An interesting question from sejasp on blocks sizes ... http://dba.ipbhost.com/index.php?showtopic=1512

A five-times improvement in i/o performance, just by changing a block size? It's the sort of thing that ought to be making the hairs on the back of your neck rise.

UPDATE: April 23rd

Well the thread has morphed into the issue of whether it is appropriate/rude/presumptious of a person to ask questions of the original poster in a forum thread.

My opinion, which I think I state pretty clearly there, is based on the following:
  • A person has asked a question in a public forum, requesting advice.
  • As their question stands, it offers the possibility of a number of different answers, and some of the answers may be inappropriate.
  • Erm ... that's it.
The most appropriate answer for the situation may be, "I do not think that any action is required here ... what you perceive as a problem does not in my opinion justify any corrective action." You know that's advice that I've given quite frequently. Here's some examples of questions to which the answer is "Do Nothing"
  1. My data warehouse application is not using bind variables ... what should I do?
  2. I have a number of indexes with more than 120 extents ... how should I reduce this?
  3. My buffer cache hit ratio has dropped below 99% to 97% ... help!
So under these circumstances, can anyone tell me what is wrong with probing for more information? Especially when you are giving free advice.

UPDATE: 24th April

Thanks for the nice thoughts there, fellas. Much appreciated.

UPDATE: 25th April

Happy ending -- see, I knew sejasp was a good bloke on account of being an Aussie. I have about nine great relatives over there (mostly poaching, but a little light larceny also).

Hey, just jokes fellas. Happy ANZAC day. (Is that appropriate? Is ANZAC day potentially a happy thing? comments welcome, as always)

6 Comments:

At 10:52 AM, Blogger Pete Scott said...

I am not too convinced that the guy has a good handle on i/o measurement and what he hopes to achieve. It may be more realistic for him to query his table (well, his development version of it) rather than a select * on ALL_OBJECTS. The bulk of his queries could just be targeted at single rows and (to my mind) bigger block size may not be a great help here.

(bet you can tell that I don't do OLTP!)

 
At 6:08 PM, Blogger David Aldridge said...

Indeed. The test is very obscure, but I also had the thought that he might be querying both the original system view and a copy of it's data, or two copies of it, otherwise this would have to be two completely different databases.

And yes, I think that a lesson on 10046's is in order down the road.

 
At 5:44 AM, Blogger Pete Scott said...

David
There is ABSOLUTELY nothing wrong with probing for more information before answering a question. At the most basic level you are validating that you understood the question; you also are demonstrating that you are THINKING about the question and your response. It also shows that you aware that there may not be a 'one-size-fits-all' or "magic bullet" answer for the question.
I would much prefer to be asked further questions to help me get to viable solution then to be 'lead up the garden path' as the expression goes.

 
At 2:26 PM, Anonymous Anonymous said...

You're just doing this for fun now, aren't you??

 
At 3:08 PM, Blogger David Aldridge said...

Um .... well I can't say that it isn't enjoyable at some level.

 
At 3:59 PM, Anonymous Anonymous said...

Amazing. It's just amazing to see what kinda moron DKB can be at times (which is like 99.99% of the time).

I see that DKB kinda skunked away after a few of the other posters jumped to your defense there.

 

Post a Comment

<< Home