Re: Table / Index Size Formula

In article <33600580.572E_at_mindspring.com>, Ed Jennings <jenningse_at_mindspring.com> writes
>I built an excel spreadsheet to calculate table/indexes sizes based on
>the information in the Oracle 7.1 SA guide. When I run existing tables
>through this spreadsheet, the figures always seem to end up about 25% on
>the high side. I triple checked the formulas, and they are exactly as
>outlined in the guide. Does anyone know if this is a growth factor
>built in to the formulas. I'd be very interested in formulas anyone
>else has come up with. In the very near future I will be calculating
>nearly 1000 tables and associated indexes. I'd very much like a more
>accurate formula.
>
>Ed

I have found the formulae to be reasonably accurate provided they are given the _average_ size of columns, rather than the maximum. If you have a development database with a reasonable sample of data and you "select avg(vsize(colname)) from table" for each column, and feed these in to the spreadsheet, I have found the results to be within 10 percent of the real figures. If I don't have sample data and have to rely on maximum sizes then I have also found 25-30 percent over, but disk space is cheap, and I would rather overestimate than underestimate.

I have also found that the index sizing is much less accurate tha table sizing. I suspect that this is because the amount of wasted space in an index depends on the insertion/deletion pattern of the data.

-- 
Jim Smith