Re: db_file_multiblock_read_count

In article <34185892.C6B5FE02_at_mail.telepac.pt>, Nuno Martins <nunomartins_at_mail.telepac.pt> wrote:
>> >Hello,
>> >if i have a db_file_multiblock_read_count = 32 then it thas not make
>> >sense having a index in tables with less than 64k . Is this true ?
>> >
 

> And there is examples where is not faster ? how did you measured that
>?
> I tried tkprof and it shows that with index still makes less physical
>reads , the elapse time in both cases is 0 .
>

That's my point. The original question was whether it makes sense to have an index on a table with less than 64, i.e. if it can be read with 1 multi-block I/O. And my answer was "it depends".
And how did I measure it. With customer satisfaction. After persuading the cost-basedf optimizer to use an index on a less-than-5-block table, the nighlty batch job is 30-40% faster. Granted, there are unusual circumstances involved. Received on Sat Sep 13 1997 - 00:00:00 CDT