Re: commits are not really commits

In general, a "commit" results in a
"redo synch write" being sent to the
log writer to write the log, and the
calling processes waits for a response
before continuing processing.

Within a pl/sql call, commits do not
result in a "redo synch write", lgwr
is posted to write asynchronously.
A "redo synch write" is posted only
when the pl/sql call completes.

This has two main side-effects
a) your processing keeps running

    when it should be waiting

b) lgwr may piggyback several of

    your commits by the time it wakes
    up and manages to do your first
    commit - so it generates less
    redo wastage, and then does
    fewer larger writes, rather than
    one small write per commit.

If you emulate lots of small transaction by writing a pl/sql loop you can easily get results which are not representative of
real activity.

--
Regards

Jonathan Lewis
http://www.jlcomp.demon.co.uk

Next Seminar dates:
(see http://www.jlcomp.demon.co.uk/seminar.html )

____Denver_______December 2/4
____England______January 21/23


The Co-operative Oracle Users' FAQ
http://www.jlcomp.demon.co.uk/faq/ind_faq.html





Martin Doherty wrote in message <1rQH9.12$Td5.59_at_news.oracle.com>...


>Connor,

>

>Fascinating comment - "... commit are not really commits...."  Can

you


>expand a little on what this means? Does this impact the logical

>(programming) view point, or do you mean differences only in the

>physical implementation?

>

>Thanks

>Martin Doherty

>

>Connor McDonald wrote:

>

>>Ed Stevens wrote:

>>

>>

>>>[snip]

>>>

>>>

>>

>>plsql may not be a good one to use for the test because commit are

not


>>really commits so you get an artificial redo rate.

>>