[Info-vax] RAID vs. MOUNT/BIND

[H Vlems]

On 17 nov, 21:45, moro... at world.std.spaamtrap.com (Michael Moroney)
wrote:
> hel... at astro.multiCLOTHESvax.de (Phillip Helbig---undress to reply) writes:
>
> >With host-based RAID, do I have the option of a) binding several small
> >shadow sets into a larger RAID 0 device or b) shadowing two (or more)
> >RAID 0 devices?  If so, which is better?
>
> The wisdom is to make stripe (RAID 0) sets from shadowsets, not shadow
> stripe sets. This is due to failure possibilities. One failed drive in a
> stripe set takes out the whole stripe set.  If you shadow two stripe sets,
> two failed drives (one in each stripe set) takes out the whole thing.  If
> you stripe shadowsets, two drive failures can still take out the whole
> thing, but only if they both are part of the same shadowset.
>
> The same logic would apply to volume sets, if you decide to go that route,
> however you actually don't have a choice since you can't shadow volume
> sets, just its members.

My comment on binding shadowsets came from this adagio: stripe
shadowsets,
never shadow stripsets, for exactly that reason.
I remember a few lectures on disk geometry (when SCSI was new to VMS)
from
a DEC engineer whose name escapes me now. He had some fascinating
examples
on how thing could go seriously wrong when that simple rule got
ignored!
Hans