RE: Table growth - disk sizing
- From: "Powell, Mark D" <mark.powell@xxxxxxx>
- To: "Oracle-L" <oracle-l@xxxxxxxxxxxxx>
- Date: Fri, 26 Aug 2005 13:15:30 -0400
For a home grown application or developed application where no real
history exists it should be possible to estimate future growth based on
expected transactions volumes if a few conditions are meant.
1- Populate a clean test system with a realistic set of starting data
using small physical space allocations
2- Run the system simulating a realistic workload distribution for a
week
3- Compare the size of the system at the end of the week to the
beginning size and divide the change in bytes by the number of
transactions.
You now have an average growth per transaction factor. But for it to be
useful you must be able to reasonably estimate the number of future
transactions and there cannot be much variation in the way the
application is used by different sites.
The problem is very few sites use realistic test data and almost no site
is capable or willing to simulate a real production workload for a full
week. Add to this the likelihood that the application either has a
great deal of variation between sites or the transaction volume
estimates are off by an order of magnitude and I would hazard a guess
that only 5% of IT shops in the world can make accurate predictions.
HTH -- Mark D Powell --
________________________________
From: oracle-l-bounce@xxxxxxxxxxxxx
[mailto:oracle-l-bounce@xxxxxxxxxxxxx] On Behalf Of Gogala, Mladen
Sent: Friday, August 26, 2005 12:52 PM
To: 'Ranko Mosic'; Gogala, Mladen
Cc: Christian.Antognini@xxxxxxxxxxxx; tomday2@xxxxxxxxx; Oracle-L
Subject: RE: Table growth - disk sizing
What I was trying to tell you is that the goal of your project can be
qualified as BS. You are trying to predict the future based on number
of transactions, instead of the historical growth information, like
everybody else. That will not work. Allow me to jovially suggest coin
toss, tea leaves, position of the stars or entrails of Oracle sales
people
as the next best methods.
The only reliable extrapolation is linear approximation, based on
historical
data. Even this gives you only short time reliability. Everything else
is BS.
--
Mladen Gogala
Ext. 121
-----Original Message-----
From: Ranko Mosic [mailto:ranko.mosic@xxxxxxxxx]
Sent: Friday, August 26, 2005 11:18 AM
To: MGogala@xxxxxxxxxxxxxxxxxxxx
Cc: Christian.Antognini@xxxxxxxxxxxx; tomday2@xxxxxxxxx; Oracle-L
Subject: Re: Table growth - disk sizing
I was looking for useful advice, not bs.
Thanks, rm.
Other related posts:
