disklabeling vs advfs volumes

Hi Managers,

This might turn out to be a stupid question, but it's a subject I don't
understand well myself.

We have a MSA1000 with an aggregate raw disk size of about 1.8 TB.  We want
to split this up into at least 3 volumes/partitions/moun-points whatever
the correct word is.  We want to keep it as a RAID-5, with spare disks for
redundancy, use AdvFS, and maximize our disk space.

The way I see it, there's many ways to split up a disk.  I could use the
MSA1000 to split the disks into three disk sets.  But obviously the more
disk "sets" I make using the MSA1000, the less storage space I will have,
because I need to dedicate one whole disk as a spare disk.

I could make one single disk set, make them into one whole Advfs domain,
and split them into many volumes.  I don't have a AdvFS Advanced Utilities
license, but I think I can obtain one.  This has the advantages of only
needing one spare disk.

Finally, I was suggested by my boss make one disk set, then split them up
using diskconfig into several disklabel logical partitions, say
/dev/disk/dsk5a, /dev/disk/dsk5b, /dev/disk/dsk5d.  Then making a single
Advfs domain, and fileset in each partition.  This sounds bizzare, but I
can't think of anything wrong with it.

My question is, is there something fundamentally flawed with the third
option?  I never really understood the concept of disklabeling partitions,
I know you use dsk*a for the root partition, dsk*b is for the swap, dsk*c
is for the entire disk - but what if I used it as a way to split it up into
3 data partitions?

--
Kevin Dea
UNIX System Administrator
Alpine Electronics Research of America