Re: Insert to Clob datatype test 8x slower on 9207 vs 9205?

From: Dave Stien <noyoudont_at_null.null.null>
Date: 17 Jul 2006 11:14:26 GMT
Message-ID: <slrnebms8l.aum.noyoudont@frodo.local.35ad.com>

Update on this following isolated testing on a little sun e250.

9207 shows lob insert 200k row benchmark taking 50% longer than 9205. I have 10min statspack snaps gathered during each run via script.

Based on the spdata below am i right in thinking the 9207 difference may be due to redo log related differences in behaviour?

(9205 does not seeem to suffer on log file parallel write and spends
more time direct path writing lob)

Thoughts, comments, suggested further tests? am i way off the mark?

cheers,
Dave.

9205 Top 5 Timed Events

Event			Waits	Time(s)	%Total elap time
direct path write (lob) 4,677   281     54.76
dbfile seq read		17,458	130	25.37
dbfile parallel write	23	46	8.93
controlfile par w 	182	42	8.28
log file parallel w	512	9	1.70

9207 Top 5 Timed Events

Event			Waits	Time(s)	%Total elap time
log file parallel w	656	204	56.39
db file sequential read	13741	99	27.28
controlfile par write	189	15	4.03
direct path write(lob)	3658	5	1.41

(9205 and 9207 databases created via same script and with init params.
Simple single disk storage and cooked ufs file system with logging One database running at a time.) Received on Mon Jul 17 2006 - 06:14:26 CDT