Re: Rule-based vs Cost-based Optimizer?
Date: Mon, 20 Mar 1995 00:25:56 GMT
Message-ID: <D5pqJ9.3M7_at_janus.border.com>
In article <acc-1903951037230001_at_pfox.magna.com.au>, acc_at_pfox.dialix.oz.au (Andrew Calvin) says:
>
>In article <3k73qj$34b_at_potogold.rmii.com>, bromberg@hobbes (Rich Bromberg)
>wrote:
>> the cost-based optimizer in a 7.0.16 and 7.1.3 ... have observed a
>number of extremely poor 'decisions' by the optimizer.
>
>I have experienced the same problem, so we have returned to rule-based
>optimisation at all times.
>
>-- Andrew Calvin -- acc_at_pfox.dialix.oz.au --
My experience is that the cost based optimiser generally does better than
rule as long as you "compute" stats. The "estimate" yeilds stats that are
too inaccurate causing the CBO. to make bad decisions.
When the CBO has good stats it generally does better, Of course there are exceptions to every rule.
Regards
GAG
Received on Mon Mar 20 1995 - 01:25:56 CET