The other suggestions so far are
all relevant. As far as writing to
the disk that the table is on,
you may be experiencing block
cleanout. If a very large number of
blocks in the table were inserted
or updated without experiencing
a fast commit, then the next read
would cause them to be updated
with commit information, thus
becoming dirty and needed a
rewrite. If this sounds like a
possibility in your case, then
a simple select count(*) from table
before creating the index may
help as this would clean all
the blocks.
--
Jonathan Lewis
Yet another Oracle-related web site: http://www.jlcomp.demon.co.uk
Adam Roze wrote in message <38C6758D.339F0603_at_royalpbk.com>...
>Hi,
>
>I am trying to create unique index on a big table.
>Table is placed on separated datafile/tablespace/disk . Monitoring disk
>activity shows that oracle reads and writes about the same
>amount of data ~2.5 MB/s.
>
>I understand that oracle has to read the table to create index but why
>does it write to it ??? Is it possible to avoid ??
>
>As the table is big ~6GB it takes an hour to create the index and I
>hope that elimination of writes would make it faster.
>
>Oracle is 7.3.4.5 on Digital Unix 4.0E
>
>Any help would be greatly appreciated.
>
>Adam
>
Received on Wed Mar 08 2000 - 00:00:00 CST
