How to create table with 1 row 1MB in size?

Question

Hello,
I am doing some R&D and want to create a Table with 1 row, which is 1 MB in size.
i.e. I want to create a row which is 1 MB in size.
I am using a 11g DB.
I do this in SQL*Plus:
(1.) CREATE TABLE onembrow  (pk NUMBER PRIMARY KEY, onembcolumn CLOB);
(2.) Since 1MB is 1024*1024 bytes (i.e. 1048576 bytes) and since in English 1 letter = 1 byte, I do this
SQL> INSERT INTO onembrow VALUES (1, RPAD('A', 1048576, 'B'));
1 row created.
(3.) Now, after committing, I do an analyze table.
SQL> ANALYZE TABLE onembrow COMPUTE STATISTICS;
Table analyzed.
(4.) Now, I check the actual size of the table using this query.
select segment_name,segment_type,bytes/1024/1024 MB
from user_segments where segment_type='TABLE' and segment_name='ONEMBROW';
SEGMENT_NAME       
SEGMENT_TYPE        
MB
ONEMBROW           
TABLE            
.0625
Why is the size only .0625 MB, when it should be 1 MB?
Here is the DB Block related parameters:
SELECT * FROM v$parameter WHERE upper(name) LIKE '%BLOCK%';
  NUM NAME                                                                                   TYPE VALUE 
  478 db_block_buffers                                                                          3 0     
  482 db_block_checksum                                                                         2 TYPICAL
  484 db_block_size                                                                             3 8192  
  682 db_file_multiblock_read_count                                                             3 128   
  942 db_block_checking                                                                         2 FALSE 
What am I doing wrong here???

Answer 1

When testing it is necessary to do something that is a reasonably realistic model of a problem you might anticipate appearing in a production system - a row of 1MB doesn't seem likely to be a useful source of information for "R&D on performance tuning"
What's wrong with creating millions of rows ?
Here's a cut and paste from a windows system running 11.2.0.3
SQL> set timing on
SQL>
SQL> drop table t1 purge;
Table dropped.
Elapsed: 00:00:00.04
SQL>
SQL> create table t1
  2  nologging
  3  as
  4  with generator as (
  5     select
  6             rownum id
  7     from dual
  8     connect by
  9             level <= 50
10  ),
11  ao as (
12     select
13             *
14     from
15             all_objects
16     where   rownum <= 50000
17  )
18  select
19     rownum          id,
20     ao.*
21  from
22     generator       v1,
23     ao
24  ;
Table created.
Elapsed: 00:00:07.09
7 seconds to generate 2.5M rows doesn't seem like a problem.  For a modelling example I have one script that generates 6.5M (carefully engineered) rows, with a couple of indexes and a foreign key or two, then collects stats (no histograms) in 3.5 minutes.
Regards
Jonathan Lewis
http://jonathanlewis.wordpress.com
Now on Twitter: @jloracle