Connor McDonald wrote:
> Oradba Linux wrote:
>
>>"Connor McDonald" <connor_mcdonald_at_yahoo.com> wrote in message
>>news:419C9CAB.1659_at_yahoo.com...
>>
>>>Oradba Linux wrote:
>>>
>>>>Connor McDonald wrote:
>>>>
>>>>>Oradba Linux wrote:
>>>>>
>>>>>
>>>>>>I have gathered statistics for all objects in our prod database.
>>>>>>How do you guys find out the next time to gather stats on the table?
>>>>>>Do you regularly put them in a schedule ? My intention is avoid
>>
>>wasting
>>
>>>>>>cpu cycles gathering statistics unless it is needed.
>>>>>
>>>>>
>>>>>check out the MONITORING clause for tables
>>>>>
>>>>>hth
>>>>>connor
>>>>
>>>>Hello Connor
>>>>
>>>>I have just ordered your book. I am liking the reviews a lot. Hopefully
>>>>i will not be disappointed.
>>>>The reason i dont want to turn on monitoring for all tables is the
>>>>overhead. Is this overhead too much? I am looking for opinions.
>>>>Also it is a datawarehouse/reporting db.
>>>>I am on oracle 9204 / sun os 32 bit
>>>
>>>There is an overhead but I'll doubt you'll notice it. Its predominantly
>>>memory based (and latch-less) with the data only being flushed down the
>>>the data dictionary at regular intervals.
>>>
>>>In fact, in v10, all tables are monitored as a default.
>>>
>>>hth
>>>connor
>>>--
>>>Connor McDonald
>>>Co-author: "Mastering Oracle PL/SQL - Practical Solutions"
>>>ISBN: 1590592174
>>>
>>>web: http://www.oracledba.co.uk
>>>web: http://www.oaktable.net
>>>email: connor_mcdonald_at_yahoo.com
>>>
>>>Coming Soon! "Oracle Insight - Tales of the OakTable"
>>>
>>>"GIVE a man a fish and he will eat for a day. But TEACH him how to fish,
>>>and...he will sit in a boat and drink beer all day"
>>>
>>>------------------------------------------------------------
>>
>>1. Will i not notice this on a heavy OLTP DB? Is this so lightweight?
>>2. My question is more what if 10% of the table which has say 60 mil rows
>>have changed, do i have to gather statistics again?
>>3. If i go do an update without actually changing the data, it will still
>>show up in dba_tab_modifications view.
>> It will not happen in a live system guaranteed. does this mean that i
>>should gather stats again
>>4. According to metalink support note <> it is says it is enough to
>>gather stats using 5% estimate. Has anybody seen this not to
>>work well.
>>I want to know how statistics gathering is handled in real live system
>>OLTP/DSS.
>>
>>Thanks in advance
>
>
> Don't get me wrong - I'm not advocating you gather stats all the time.
> I just said that monitoring should not have a noticeable impact.
>
> My preference is to only calculate stats when things go bad - why would
> you do so otherwise ?
>
> hth
> connor
Hello Connor
Started reading your book. Liking it so far.
Doesn't calculating stats when there is a problem becomes a reactive
approach. Should a problem arise in the middle of the day then running
this gather stats which is cpu intensive will be a problem.
Received on Fri Nov 19 2004 - 12:10:09 CST
