Re: Extent Management in SYSTEM tablespace

Connor McDonald wrote:
> Joel Garry wrote:
> >
> > >DMT is dead...
> >
> > And like fish and visiting relatives, should be thrown out after 3
> > days. (Finally convinced a stick-in-the-mud customer to let me get
rid
> > of them this past weekend.)
> >
> > Did anyone ever come up with a reference or refute for the
statement
> > that all the LMT extent sizes internally use 64K extents despite
what
> > oracle reports?
> >
> > jg
> > --
> > @home.com is bogus.
> > "My head is *spinning*...but I am willing to bet my savings right
now
> > that the next reincarnation of HTMLDB is gonna end up in the next
AS
> > like
> > WebDB...Portal...Anyone wanna bet against me ? :) " - Robert from
CT
> > "I will." - Tom Kyte
>
> For uniform you'll see 1 bit per extent size nominated in the
datafile
> dumps. For auto, you'll see 1 bit per 64k
>
> there's an article on www.oaktable.net of mine about it

That article definitely should be in the faq for LMT.

What struck me strange was I just changed an export to DIRECT=Y (9206) and discovered the bit about recordlength being limited to 64K-1. So I guess it is only 65535/65536 as fast as it could be, unless such a limit means having to do a lot of contortions to deal with 64K of data at a time, ie "hang on to that 8th block cause I could only fit seven into 64K-1," if those are 8K blocks with db_file_multiblock_read_count set.

I'm sure Oracle's answer would be "datapump." :-)

jg

--
@home.com is bogus.
Fuzzy Grey Copyright Law:
http://www.informationweek.com/story/showArticle.jhtml;jsessionid=4LIZMGIVEVVQ4QSNDBGCKHSCJUMEKJVN?articleID=59100462