agarwalp_at_eeism.com wrote:
> Hi,
>
> Following is the excerpt from an article on Transaction logs under
> Checkpoints Heading by Peter Gulutzan
> (http://www.dbazine.com/db2/db2-disarticles/gulutzan2)
>
> "Users need recoveries to work faster, so there are times when the DBMS
> will write absolutely everything. This enables it to flush all dirty
> data pages to the database too. Then the DBMS writes a log record
> marking that a "checkpoint" has been done. If a crash occurs later,
> recovery is easy because the DBMS only needs to roll forward with log
> records since the last checkpoint"
>
> This is somewhat confusing me. What i interpret is that when the DBMS
> writes dirty data records to the datafiles during a check point, it
> updqate the log file saying that a check point is done. So my question
> is why & what does it have to roll forward, even though i have the data
> on the physical datafile.
I haven't read the article but what comes through pretty clearly from
the URL is that it refers to DB2: Not Oracle.
--
Daniel A. Morgan
http://www.psoug.org
damorgan_at_x.washington.edu
(replace x with u to respond)
Received on Wed Jun 15 2005 - 09:00:55 CDT
