Re: Myth revisited ...

"Niall Litchfield" <n-litchfield_at_audit-commission.gov.uk> wrote in message news:3fb9f03d$0$265$ed9e5944_at_reading.news.pipex.net...

> between. And assuming that Nuno is correct about the performance (I haven't
> tested) it probably depends on the nature of the data access. I guess I was

Absolutely, nature and pattern of access is important. What I noted with "out of line" (lack of a better term?) storage is it's not guaranteed to be contiguous or as easy to find as "inline" storage. After all, inline is just a stroll down the block while the other means another index follow-up and another block(s) read/written. Of course it will be much slower, assuming same size of LOB. Once sizes are different (inline small, outofline large), then times compound even more for the outofline access.

The little experience I have with this is with CLOBs. It confirms this. Inline ones have virtually no diff from varchar for normal get/put access. Out of line are considerably slower even if size difference is just borderline. This goes for both index access of base row or FTS.

-- 
Cheers
Nuno Souto
wizofoz2k_at_yahoo.com.au.nospam