Re: Recovery considerations of checkpoint vs. log switch

I don't forsee any problem with this method. Checkpointing every hour will force DBWR to write all changed blocks to the disk and this will speed up your recovery in case of a crash. Now in case you crash your database along with the online redo logs then you could lose data. But what are the chances of that happening ? You have to offset it with the perfomance gain you get.

Denny

In article <7at5ot$p2u$1_at_tortue.nbnet.nb.ca>,   gosh_at_canada.com wrote:
> I have a system that runs an online application (7.3.3 + middleware)
> during the day and a batch run at night. The redo logs are sized to
> switch about once every 1 - 2 hours during the day. During batch that
> frequency can reach one every 2 - 3 minutes. Needless to say I need
> to do something. The something we've decided on (but talk me out of
> it if you want) is to increase the redo log size to try to get a redo
> log switch once every half hour during the batch run. During
> the day we will force a checkpoint every hour, so as to be able to recover
> if the machine dies.
>
> Given the above plan, is there any recovery advantage to forcing a log switch
> every hour over using a timed checkpoint strategy?
>
> Any comments would be appreciated.
>
> Thanks,
> Casper
> --
> Paul Schenk gosh_at_canada.com
> Verbing weirds language. - Calvin
>

Denny Koovakattu

-----------== Posted via Deja News, The Discussion Network ==---------- http://www.dejanews.com/ Search, Read, Discuss, or Start Your Own Received on Tue Feb 23 1999 - 19:56:52 CST