Re: Index management

Howard,

Snip: > Er, that would be "Howard", not "Roger".

So sorry Howard, won't happen again.

Snip: > ...you make your first mistake, since the advice should be 'not to
> unless', not 'rebuild unless'.

Again, you cut the sentence short only quoting enough to "prove" your point which is no proof at all.

Snip: That sentence doesn't even make sense to me. "I tell them to compare it
> [what "it" are we talking about? From the context, it sounds like it
> should be "the number of leaf nodes of the index"]

Again, in the original context, the presentation in which I was quoted, incompletely, the context was clustering factors. Sorry I thought you knew this. But I should perhaps have requoted myself to those who didn't back track to the presentation.

Snip: If you don't like being
> quoted, then don't say it in the first place.

Advice we should all follow, however, delibrately misleading quotes that don't take into account supporting material stray in to dangerous territory, especially when they are used to damage the reputations of others.

Sinp: If your published works told the facts of the
> matter, that an index rebuild hardly ever makes a difference, but that
> there are always exceptions, then I doubt anyone would be quoting you
> otherwise.

If people look at the recommendations in total, then they know I only advocate rebuilds when it is statistically indicated or performance points at an index that is showing the signs of aging I have discussed. In the situations I discuss rebuilds have reduced processing times from 5 hours to 2 hours, reduced index sizes from 80 gig to 20-30 gig and in another case reduced index tablespace utilization by 90%. In testing changing the column orders in a concatenated index can reduce cluetering factors by an order of magnitude. I could go on, but it is probably casting pearls.

Mike Received on Thu May 06 2004 - 08:28:57 CDT