Re: Frequent Checkpointing ...

From: Tanel Poder <change_to_my_first_name_at_integrid.info>
Date: Thu, 18 Sep 2003 23:32:29 +0300
Message-ID: <3f6a165d$1_1@news.estpak.ee>

Starting from 8i you have a better option, fast_start_io_target and in 9i fast_start_mttr_target which specify max IO target Oracle has to do for crash recovery. This is called incremental checkpointing (if I recall correctly).

But before that, I've heard people doing more frequent checkpoints (which means higher overall IO) or making their redologs so big, that they'd switch less than once per day -> but the logswitch was issued manually during low activity time. This had other problems -> logfile sizes, both from storage and recovery viewpoint...

Tanel.

"Domenic G." <domenicg_at_hotmail.com> wrote in message news:c7e08a19.0309180734.2e09803d_at_posting.google.com...
> I was just reading the INDEX/TABLE separation thread -- how about the
> checkpoint myth? From my experience, it is better to checkpoint
> frequently and keep those disks busy at an even rate vs. delaying
> checkpoints for 20 minutes and then all of a sudden you've got a
> workload surge to disk. To me that never made sense. Checkpoint
> frequently, every few minutes -- keep the checkpoint workload even and
> steady.
>
> Domenic.
Received on Thu Sep 18 2003 - 15:32:29 CDT