Oracle FAQ | Your Portal to the Oracle Knowledge Grid |
Home -> Community -> Usenet -> c.d.o.misc -> Re: Help on execution plan
> might be worthwhile extending the index to include all columns that need
> to be returned (a valid tuning trick)
Yes I'm already evaluating that, to check if disk space is OK in the production environment
> how many rows are you returning?
As I said: about 3500 (out of 4.5 million)
Thanks for your input
Thomas
mcstock schrieb:
> might be worthwhile extending the index to include all columns that need to
> be returned (a valid tuning trick)
>
> how many rows are you returning?
>
> "Thomas Kellerer" <spam_eater_at_gmx.net> wrote in message
> news:bodhot$1c6q3a$1_at_ID-13919.news.uni-berlin.de...
>
>>Mark, >> >>thanks for your feedback. I was thinking about the columns beeing part of
>>index already. The funny thing about this, is that the index which *is*
>>when I remove the two columns is not the index that is beeing reported by >>EXPLAIN PLAN. The index beeing use is the one over (id, product_no)
>>there is one over (id, name, product_no, ca_code). That's basically the
>>why I wasn't considering the index read. >> >>And yes, I have tried using hints, but any of the indexes I specify make
>>doesn't change too much. >> >>I am aware of the fact that an index scan *can* be more expensive then a
>>taking into account that the table has 4.5 million rows (in the test >>environment, production has about 9 million rows) I would expect an index
>>scan still beeing less expensive.... >> >>Thomas >> >> >>mcstock schrieb: >> >> >>>the columns that remain in the select list likely are being read from
>>>index -- is there any other type table access for 'big_table' in the >>>execution plan replacing the FTS? >>>does anything else in the plan change? >>>have you looked into hints to force the index range scan (which may not
>>>so cheap with the additional columns included -- the optimizer may be
>>>a correct choice in this case) >>> >>