This discussion reminds me of the first DBA course I attended back in 1989 I
think. Not for Oracle but for IBM SQL/DS (the forerunner to DB2)
At a time when disk space really was at a premium (we had just bought 21GiB,
2nd user, for £500,000), part of the course was sizing, and we sat doing lots
of careful calculations to get exact row sizes - taking into account length
bytes and all that fiddly stuff. Really detailed, getting a very carfully
considered size for tables and indexes, and the instructor said
"OK, now take what you've got and double it. you always run out of space
quicker than you think" - spend hours calculating exact sizes and then add 100%.
I've sized systems the same way ever since. Work out what you need. Double it.
Convince someone they need to pay for it.
From: oracle-l-bounce@xxxxxxxxxxxxx <oracle-l-bounce@xxxxxxxxxxxxx> on behalf
of Jonathan Lewis <jlewisoracle@xxxxxxxxx>
Sent: 22 February 2021 15:42
To: Pap <oracle.developer35@xxxxxxxxx>
Cc: Lothar Flatz <l.flatz@xxxxxxxxxx>; jack@xxxxxxxxxxxx <jack@xxxxxxxxxxxx>;
Mark W. Farnham <mwf@xxxxxxxx>; Oracle L <oracle-l@xxxxxxxxxxxxx>
Subject: Re: Size estimation
A couple of points to watch out for - again related to a small model for a big
Are you counting leaf blocks in the index to get 4MB or looking at the size of
the segment ?