On 17 May 96 at 7:22, Michael L Gueterman wrote:
 
> On May 17 I wrote:
>
>> The stats are not modified each time you add/delete a record or change
>> an index value.  Therefore, the stat information is typically behind
>> the data.  The more dynamic the data, the more you should update stats
>> on that table.
>
> That's not necessarily true, if you have a VERY dynamic table that
> you "know" the best access path for (and your not overiding the
> optimizer :), you can cause a tremendous amount of heartburn by
> doing an Update Statistics and giving the optimizer a 'clear view
> of the world'.  It may then decide that your previous access path
> isn't the best and start giving you back another way.
 
Michael makes a great point here, that I forgot.  That is, if you like
the response you're getting now, don't do the update stats.  The update
stats could, actually, reduce your response time.  Since the optimizer
is only a mathematical algorithm, it is based on information *about*
your data, not based *on* your data.  This means that in some cases,
depending on your data, the optimizer could retrieve more slowly than
if a different index was used.
 
The ALLBASE optimizer is one of the best, if not the best, in the
industry, which is one of the reasons it tops the speed of others.
However, all optimizers will have problems with corner cases.  This is
similar to the hashing algorithm within IMAGE.  Generally, it's great,
but there are corner cases where you could be in trouble.
 
 
Larry Boyd    <[log in to unmask]>
"Each problem solved creates the opportunity to solve the next problem
          that the last solution created." - Richard Pascale
(These opinions are my own and not those of Hewlett-Packard.)