I've benchmarked a few scenarios to load mysqldumped data into an
innobase-table (about 27 columns, mostly int's and float's) with about
150.000 rows with primary key and an additional unique key. The mysqldumped
data was generated one insert per row (not dumped with --extended-inserts).
1. cat test.sql | mysql innobase => 306 seconds
2. cat test.sql | mysql innobase, but with autocommit=0 and all inserts
between BEGIN/COMMIT => 124 seconds
3. same as 1. but with index creation after the inserts. => 264 seconds +
additional 59 seconds for index recreation.
4. same as 2. but with index creation after the inserts. => 75 seconds +
additional 59 seconds for index recreation.
I noticed that in 3. and 4. the machine load was constantly at 100% during
the inserts.
but during index recreation the load springs wildley between 0 and 100%, so
the machine is not maxed out in terms of CPU utilization here (in other
words, the machine is waiting for something to do :-). The same applies to
2. when commit() was send.
I tested also 1. and 3. with a myisam-table. Index recreation draws
constantly 100% load from the machine here, therfore it's faster.
results:
5. same as 1. but with myisam type => 116 seconds
6. same as 3. but with myisam type => 65 seconds + additional 33 seconds for
index recreation.
system is a celeron 466 / 66MHz FSB, 512 MB RAM, Linux 2.4.2, mysql-3.23.35
If the numbers are not useful to anybody, think of this posting as spam and
forget myself...
--Andy
/"\
\ / ASCII RIBBON CAMPAIGN
X AGAINST HTML EMAIL
/ \ AND POSTINGS
---------------------------------------------------------------------
Before posting, please check:
http://www.mysql.com/manual.php (the manual)
http://lists.mysql.com/ (the list archive)
To request this thread, e-mail <[EMAIL PROTECTED]>
To unsubscribe, e-mail <[EMAIL PROTECTED]>
Trouble unsubscribing? Try: http://lists.mysql.com/php/unsubscribe.php
