Andrew Wheadon writes:
> Hi everybody,
> 	I'm currently adding diskspace to my machine. Unluckily
> 	the 'ftp'-section will be too large to fit on the largest
> 	disk soon and since I don't want to keep having to move
> 	'.a', '.b' etc. around to accomodate changes in the structure
> 	I'll be using 'ccd'. (Assuming nobody tells me it's instable)

On a 386dx40 with 8MB RAM and 2 x DEC 1GB SCSI1 disks, AHA1542CF SCSI
controller, running NetBSD-1.1, my ftp server has been running for ~190
days without hiccups. The configuration of the disks is:
	part	size	mount		part	size	mount
	----	----	-----		----	----	-----
	sd0a	128	/		sd1a	128	/var
	sd0b	64	swap		sd1b	64	swap
	sd0e	800	(ccd)		sd1e	800	(ccd)

sd0e & sd1e are interleaved at 16K.

> 	How much faster will two identical disks be when
> 	using 'interleaved' in comparison to 'serial' with ccd ?
> 	These two identical disks are according to 'dmesg'
> 	different in that one is 'SCSI2' while the other is 'SCSI1'
> 	Will mixing the two on the same bus cause 
> 	(performance/other) problems ?

>From my experience, ccd is stable. I've found that on slower disks
it's not a problem to run two or three on the same bus, especially
with the ncr controller, especially if you interleave.

My only hassle with ccd is that when you are playing around with
parameters on ccd's, you can't re-use a ccd (e.g,. ccd0) until reboot.
I can't remember the specifics (hazy memory), but even when you
unconfig the device you can't reuse the same device. I never thought
much of it (you only get bitten when you "ccdconfig, newfs, bonnie
ccdconfig -U", to determine the optimal layout). I can attempt to
reproduce if people are interested.

--
Luke Mewburn                            UNIX Technical Support
<luke.mewburn@itg.telstra.com.au>     CPR Project, ITG, Telstra.
Phone: +61 3 9634 2112