Jonathan Lewis wrote:
> "Dereck L. Dietz" <dietzdl_at_ameritech.net> wrote in message
> news:qa2ah.254$Py2.90_at_newssvr27.news.prodigy.net...
>> In reading a study guide for Oracle 9i it described the data block
>> headers.
>> One item it mentions is the Row Directory which it states occupies (4 *
>> the
>> number of rows) bytes.
>>
>> The other book I used to get the formula I have been using to
>> "guestimate"
>> table/index sizing was the Oracle 8 DBA Manual and it just had the block
>> overhead as 90 for tables and 161 for indexes.
>>
>> My question is probably a simple one and I apologize if it is but when it
>> states (4 * the number of rows) am I correct in thinking it is referring
>> to
>> the number of rows in the data BLOCK and not the entire table?
>>
>> Another question is, isn't this a Catch-22 situation? It seems that in
>> order to know what the block overhead is you need to know how many rows
>> will
>> be in the block; but in order to know how many rows will fit in the
>> block
>> you need to know the free bytes in the block (block size minus the block
>> overhead).
>>
>> Thanks.
>>
>>
>
>
> It's open to argument whether you should label
> the row directory as part of the block header -
> and it looks as if your second source has not done so.
>
> The row directory is 2 bytes per row.
>
> For tables, there is also a 3-byte per row overhead
> in-line with the row; and for indexes where is a 2-byte
> row overhead in-line with the row. Possibly somewhere
> in the row-overheads your first source has managed to
> invent the 4 bytes for the row-directory.
And if you build the table using ROWDEPENDENCIES there is another
6 bytes per row required.
--
Daniel A. Morgan
University of Washington
damorgan_at_x.washington.edu
(replace x with u to respond)
Puget Sound Oracle Users Group
www.psoug.org
Received on Mon Nov 27 2006 - 10:38:33 CST
