Re: Questions: Log interval checkpoints and log switches

>1) Why are my log switches occuring so frequently (given the size of my
redo
>log files)?

Your log_checkpoint_interval is the default 10000 OS blocks. Assuming an OS block size of 2048 - this would be 19.5M. A checkpoint should occur when 19.5M of your log file gets filled up. If you want to have checkpoints only at log switch increase the log_checkpoint_interval such that it is slightly more than your log file size. Be warned that this will increase recovery time and/or amount of data lost in case of db failure.

2) Why are my interval checkpoints occuring so frequently between
>log switches?

See above

3) Does inserting 150 Mb into tables with no logging specified
>result in redo log entries?

4) Does inserting 200 Mb into tables with logging
>specified result in 200 Mb of redo log entries?
More redo than the amount of data manipulated is generated becuase the database also has to do its own processing. That is space allocation, management, log switching etc.

5) When is the log buffer
>flushed to disk? When it reaches 1/3 full?
I myself would like a definitive answer to this. Someone please....

6) When are interval checkpoints
>triggered? Is the log_checkpoint_interval parameter multiplied by the
>db_block_size parameter to determine when this occurs?
It is the number of OS blocks modified

6) What should I
>change to reduce the number of interval checkpoints?
Look at changing your log_checkpoint_interval parameter

7) Any other
>recommendations to improve overall load time?Thank you.
Use DIRECT path load whenever possible. make sure dynamic space allocation/management does not happen by sizing your extents right. My loads of 250M and above take about 5minutes with DIRECT and UNRECOVERABLE options and my server is smaller than yours.

HTH
Venkat Received on Wed May 20 1998 - 10:21:37 CDT