Reading the SCN from a block that is in memory is
as tremendously complex as following a pointer to a location
in memory. I am soooooo concerned about the performance
implications of that! I'm sure so is Oracle...
Reading the SCN from a block that is not in memory is
as inefficient as ANY other I/O to retrieve a database page
into cache, in ANY database software. Again, a major concern
that Oracle and the world will lose a lot of sleep over...
--
Cheers
Nuno Souto
wizofoz2k_at_yahoo.com.au.nospam
"rkusenet" <rkusenet_at_sympatico.ca> wrote in message
news:c4d9n3$2h3mqp$1_at_ID-75254.news.uni-berlin.de...
> Not only that. Oracle's versioning is forced even when there is no
> need. I proved this to Daniel few months back. A simple query on yesterday's
> data (which today is only in read mode) can be answered in
informix/db2/sqlserver
> without any locks (READ UNCOMMITTED), giving best possible performance. In
Oracle
> the system has to still read the SCN from the block to determine the version.
> However efficient it may be, it is still not the same.
Received on Wed Mar 31 2004 - 05:21:24 CST
