HP3000-L Archives

January 2000, Week 4

HP3000-L@RAVEN.UTC.EDU

Options: Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
Steve Dirickson <[log in to unmask]>
Reply To:
Steve Dirickson <[log in to unmask]>
Date:
Fri, 28 Jan 2000 20:22:52 -0800
Content-Type:
text/plain
Parts/Attachments:
text/plain (37 lines)
> -----Original Message-----
> From: Michael Hurdle
> Subject: Binding Model 20 Mechs - Disc Loss
>
>
> Has anybody out there had success in binding 4 or 6 mechs in
> a Model 20 into
> one LUN (Raid-5) on an MPE system, and if so, how much disc
> loss is incurred
> due to the parity/striping? I had always thought you either
> bound them in a
> mirrored pair (2 mechs), or Raid-5 (3 or 5 mechs). Now I am
> being told by
> one individual in the HP Response Center that you can even
> bind up to 20
> mechs in Raid-5 -- and I'm having a bit of hard time
> accepting that. Please
> share your experiences by e-mail.

There's nothing magic about the numbers 3 or 5; a RAID 5 array can have
whatever number of spindles you want (as long as you don't want 2--the
minimum is 3). The absolute overhead of RAID 5 is the same no matter how
many spindles are in the array: 1 spindle. The overhead as a percentage of
array size goes down as you add spindles. RAID 5 arrays up to 8 spindles
seem to be the most common, though I'm sure there are arrays in use with
many more than that.

Also, note that the term "spindle" as used above may actually be a compound
entity. For example, you could have an array (sometimes known as "RAID 50"
or "RAID 0+5" or something similar) that is a RAID 5 array composed of 7
"logical spindles", each of which is actually a non-parity stripe set
containing anywhere from 2 to 20 (or more) physical spindles.


Steve Dirickson   WestWin Consulting
[log in to unmask]   (360) 598-6111

ATOM RSS1 RSS2