Re: The Index Clustering Factor Explained

Noons wrote:
> "Howard J. Rogers" <hjr_at_dizwell.com> wrote in message
> news:40a04df7$0$12393$afc38c87_at_news.optusnet.com.au...
>
>

>>www.dizwell.com  -> FAQ -> Performance Tuning.
>>
>>It *is* a bit of a rush-job, so I'm open to corrections and suggestions.

Not sure what you mean. The indexes used in the actual worked examples were non-unique (in declaration, if not in content).

And the Clustering Factor calculation would not change any (ie, *how* it is calculated: if I follow this index entry, what table block do I have to visit...Oh, stone me it's the same as the last visit I made, therefore I won't count it etc etc).

> The cf would be even more relevant there as it should tell us how many
> blocks may be visited when retrieving rows of a given value of the key.

Well, that's where you start having to consider other statistics in DBA_INDEXES, such as DISTINCT_KEYS and AVG_DATA_BLOCKS_PER_KEY and so on, isn't it? In which case, it would become a "What's the DBA_INDEXES view" paper, not a 'what's the clustering factor' one. Or a "tell me everything about indexes" paper!

> If the rows are adjacent on the blocks for a given value of the key,
> then it is a "good" cf. If the opposite (one row/block for all rows of
> a given key value) then the order of rows in the table may cause problems
> for retrieving all rows of a given key value.
>
> That sort of thing.

Mmmm. Not sure it would make any difference to the paper as written or intended, would it? I mean, I'm not saying the other statistics are not interesting, but they'd be outside of scope for *that* paper. I think.

Or maybe I've just missed the point.

How about mailing me?

Regards
HJR Received on Tue May 11 2004 - 07:18:30 CDT