A late entry to this post, but following up Stefan's comment about memory and buffering, here's a highlight from SQL*Plus "Text_content" is a CLOB of about 120,000 bytes. I've deleted some of the lines from the output SQL> set autotrace traceonly statistics SQL> set timing on SQL> set longchunksize 131072 SQL> select text_content from test_lobs where id = 1; Elapsed: 00:00:00.00 Statistics ---------------------------------------------------------- 25 consistent gets 16 physical reads 120945 bytes sent via SQL*Net to client 792 bytes received via SQL*Net from client 5 SQL*Net roundtrips to/from client 1 rows processed SQL> set longchunksize 8192 SQL> select text_content from test_lobs where id = 1; Statistics ---------------------------------------------------------- 71 consistent gets 29 physical reads 123400 bytes sent via SQL*Net to client 2625 bytes received via SQL*Net from client 18 SQL*Net roundtrips to/from client 1 rows processed SQL> set longchunksize 80 SQL> select text_content from test_lobs where id = 1; Statistics ---------------------------------------------------------- 5181 consistent gets 1508 physical reads 402477 bytes sent via SQL*Net to client 211164 bytes received via SQL*Net from client 1497 SQL*Net roundtrips to/from client 1 rows processed The number of roundtrips is affected by the buffer size set of the lob (which is controlled by the longchunksize parameter in SQL*Plus). The impact on buffer gets isalso highly visible. The effect on the physical reds is dependent on whether the LOB is declared cache or
nocache. Regards Jonathan Lewis http://jonathanlewis.wordpress.com Author: Cost Based Oracle: Fundamentals http://www.jlcomp.demon.co.uk/cbo_book/ind_book.html The Co-operative Oracle Users' FAQ http://www.jlcomp.demon.co.uk/faq/ind_faq.html----- Original Message -----
The query that I am working on retrieves 67 rows each of which has one blob column. It takes a total of 11 seconds. - Peter Schauss
-- //www.freelists.org/webpage/oracle-l