Re: Redo logs a Quick I/O enabled file system

On Sep 13, 4:04 pm, Mike <n00s..._at_comcast.net> wrote:
> On Sep 13, 3:38 pm, "fitzjarr..._at_cox.net" <fitzjarr..._at_cox.net> wrote:
>
> > On Sep 13, 1:35 pm, Mike <n00s..._at_comcast.net> wrote:
>
> > > What is the impact of placing the Oracle redo logs on a Veritas Quick
> > > I/O enabled file system?
>
> > > --Thanks,
> > > --Mike Jr.
>
> > No 'impact', really, but no benefit, either, as they are written
> > serially. I found no reason to place redo logs on such a file system.
>
> David,
> Thank you. I should have been more specific and said "Cached Quick I/
> O". I know that for a UFS file system Direct I/O should be specified
> for the redo logs to maximize performance. UFS caching can kill
> performance.
>
> I am looking at a system where the redo logs have been placed on a
> VxFS with "Cached Quick I/O" enabled. Will this impact performance
> the same way that cached UFS would? Or is Veritas smarter than this?
> I am seeing I/O contention (vmstat showing at least one process
> blocked and CPU > 80% idle; iostat shows the device containing the
> redo logs at > 60% busy). I was wondering if moving the redo logs to
> a UFS Direct I/O file system would improve things.

Doing more research, it might be easier to use $ qioadmin -S filename=off mount_point

to disable caching on the individual redo log files rather than moving them.

>
> Thanks again,
> Mike Jr.
>
>
>
>
>
> > David Fitzjarrell- Hide quoted text -
>
> - Show quoted text -
Received on Thu Sep 13 2007 - 18:48:53 CDT