Re: HDD vs. RAID (was Re: Lilo Q)
On Mon, 2002-06-10 at 13:39, Dave Sherohman wrote:
> On Mon, Jun 10, 2002 at 12:07:22PM -0500, Ron Johnson wrote:
> > The problem with JBODs (just big ole disks, i.e. single disks)
>
> JBOD = Just a Bunch Of Disks, i.e., several drives operating
> independently. A JBOD can be organized into a RAID, but doesn't have
> to be.
We must work in different shops...
> > With RAID solutions, the read-write heads
> > will be in as many different places at once as you have disks.
>
> This is primarily a benefit in RAID0 or 5 configurations. RAID1 could
> benefit from it also, but a lot of RAID implementations are too stupid
> to take advantage of it. RAID4 loses some of this benefit due to the
> limitations of having all the parity data on a single disk.
If I remember correctly from the last time I created a RAID set,
the docs said:
WRITE: READ
Fastest RAID1+0 RAID1+0
RAID0 RAID0
single RAID5
RAID1 RAID4
RAID4 RAID1
Slowest RAID5 single
> > Note, though, that since the CPU overhead from calculating RAID[45]
> > recovery blocks necessitates a caching controller. Otherwise,
> > write speeds will be slower.
>
> RAID1 is also typically slower since the write isn't considered to be
> complete until it has taken place on all disks (having read-write heads
> in many places helps reads and hurts writes).
In the last few years, when speed is imperative, we've bitten
the cost bullet and gone with RAID1+0, mirrored stripesets (or
striped mirrorsets; I don't know how the controller internally
handles it).
--
+---------------------------------------------------------+
| Ron Johnson, Jr. Home: ron.l.johnson@cox.net |
| Jefferson, LA USA http://ronandheather.dhs.org:81 |
| |
| "I have created a government of whirled peas..." |
| Maharishi Mahesh Yogi, 12-May-2002, |
! CNN, Larry King Live |
+---------------------------------------------------------+
--
To UNSUBSCRIBE, email to debian-user-request@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Reply to: