On 06/30/15 04:31, Robert Elz wrote:
Date: Mon, 29 Jun 2015 22:54:19 -0453
From: "William A. Mahaffey III" <w...@hiwaay.net>
Message-ID: <55921174.3020...@hiwaay.net>
| Which is right, the list & I or those lying man pages ?
The list, and the man pages, both say the same thing as I recall.
For any stripe on the raid, N-1 drives hold the data, and one drive
holds the parity.
Whether the same drive holds all the parity. or whether it is spread
around the drives was soething I said I wasn't sure about (in a previous
message on the list - relating to NetBSD's implementation).
To be clear, I *wasn't* casting aspersions, just trying to be flip,
sorry for any offense !!!
But why do you care? Unless you're planning on working on the raidframe
implementation, the only thing it affects is performance. (Spreading the
parity around allows all drives to participate in normal read operations,
and the extra writes on write ops to be spread around all of the drives,
instead of leaving the parity drive idle on read ops, and a bottleneck on
writes).
Aside from that (and if you are seeking truly high performance, raid5 is not
the solution) it makes no difference at all to anything that you do, or
observe.
kre
I am seeking highest storage density/usage with more reliability than
RAID0, so RAID5 (or possibly RAID0 w/ another machine backing up) looks
like the best option. RAID10 would probably be more reliable, but 1/2
storage usage. With only 6 drives, or more accurately slices, I want as
many active as possible. RAID5 would have 4 active (w/ 1 spare, 1
parity) & some fault tolerance, RAID0 would have all 6 active, but no
fault tolerance.
--
William A. Mahaffey III
----------------------------------------------------------------------
"The M1 Garand is without doubt the finest implement of war
ever devised by man."
-- Gen. George S. Patton Jr.