Subject: Re: lost interrupt, device timeout...
To: Jukka Salmi <j+nbsd@2004.salmi.ch>
From: David Brownlee <abs@NetBSD.org>
List: netbsd-help
Date: 10/26/2004 12:08:57
On Tue, 26 Oct 2004, Jukka Salmi wrote:
> David Brownlee --> netbsd-help (2004-10-26 10:41:54 +0100):
>> I've seen this behaviour (raidframe across two IDE disks on separate
>> channels gives timeouts), on most of my raidframe machines.
>
> The machine in question uses RAIDframe (level 1) as well.
>
>
>> I've seen this at least on piixide and rccide, less so in viaide.
>
> So viaide and pdcide join the club...
>
> IIUC this is not critical, is it? Just suboptimal, maybe... What is
> considered better: to connect the disks to the same channel, thus
> degrading performance, or to connect to seperate channels and ignore
> the timout messages?
Benchmark both to see, maybe via pkgsrc/benchmarks/bonnie++
My guess is the same channel will be fractionally slower
normally, but the timeouts will _really_ hurt performance.
--
David/absolute -- www.NetBSD.org: No hype required --