Posted by Robert Klemme on 10/26/05 12:08

Erland Sommarskog wrote:
> (gilles27@talk21.com) writes:
>> Our review of the primary keys was triggered by one of our customers
>> complaining about the performance of the database. They claimed to
>> have "re-sequenced the primary key which reduced Index reads by a
>> factor of 5". Do you have any idea as to how they arrived at this
>> conclusion?
>
> Sounds like hogwash to me.

I don't know the table structure and queries but there actually are
scenarios with different queries against the same table where the overall
benefit of reordering columns is significant.

> But if you are an evil man, please ask them
> for a repro that demonstrates this. (For the sake of customer
> relations, you might prefer to avoid it though. :-)

That depends. From my experience customers actually honor showing
interest in their findings. Remember that they experienced a success
because they were able to implement such a dramatic improvement. In the
end both sides benefit which in turn can improve customer satisfaction.

> Gert-Jan's theory about fragmentation being fixed by reindexing sounds
> plausible to me.

Yes, that's another option.

Kind regards

robert

[Back to original message]