I was asked to comment on the following blog post and question:
http://news.com.com/8301-10784_3-9740476-7.html?tag=bl
Quote:
One point in particular: if a raid level zero controller fails in an
external hard disk, its pretty much fatal. I've spoken to Ontrack and LaCie
about this. If however, a raid zero controller is a PCI card on the
motherboard and easily swapped, can the data be recovered (assuming the
disks themselves are OK)?
I have cautioned people about "double-drives" like some of the LaCie units
ever since they first appeared. Their RAID 0 internal design (and resultant
shortcomings) were immediately obvious not only due to the size of the
units (big enough to fit two 3.5" drives internally), but especially due to
the fact that they were being offered as xxxGB "drives" when I *knew* that
drives of only half that capacity were available on the market.
Unfortunately while their designs and shortcomings are obvious to a
professional, virtually none of the published product reviews I've seen
point out the multiple internal drive/RAID 0 configuration or the potential
ramifications.
However, while these devices may have some shortcomings, if the internal
controller or other circuitry fails, it *should* or at least *might* be
possible to swap the otherwise standard internal drives over to another
identical unit in order to restore array functionality (recover the data).
This scenario could be easily tested by purchasing two identical units,
loading them with different data, and then simply swapping the drives
between them. If the data were readable after the swap, then a similar
recovery after failure should be possible as well. Other factors may be
involved, for example the donor unit may need to have the same internal
circuit or firmware revision for this to work, meaning that units which
seem identical on the exterior may not be identical at all.
Likewise it should be possible to replace a failed RAID controller (or
motherboard with integrated RAID controller) with an identical unit,
recovering the array (and data) as well.
But recovery is a last, last resort, the main issue here is backup: While I
don't normally recommend RAID 0 for data storage due to the increased risk
of failure, that risk is essentially immaterial with respect to data loss.
Perhaps a better title for the entry might be: "Don't get burned by a
failure to backup", as those clients were not "burned by their inadvertent
use of RAID 0", instead they were burned by their *advertent* failure to
backup their data! They apparently failed to understand that ALL data that
is not backed up is subject to instantaneous and total loss.
They are unfortunately not alone. Few individuals seem to understand the
need for backup, which is becoming even more of a problem in these days of
terabyte hard drives. Years ago I was a proponent of tape backup, as
multiple tapes would allow one to easily and inexpensively make the
requisite minimum two backups of important data (with one of them ideally
stored off-site). But that was in the days when the capacity of tape
storage maintained parity with the capacity of disk storage.
Today the only thing I know of that can easily and inexpensively backup a
1TB drive is *another* 1TB drive.
And of course one backup isn't enough, as in that case when a failure
occurs you would be running without backup from the time of the failure
until the repair and restore were completed. Any subsequent failure during
*that* time could mean total loss. This means at a minimum that a second
backup is required in order to be protected over the duration of a storage
failure.
Bottom line: RAID at any level is no substitute for backup. I recommend
storage volumes or devices be acquired in *triplicate* (at a minimum). That
means if you use a device (or array) containing a volume with xxxGB of
capacity, then you should insure that you have at least two additional
separate physical devices containing volumes with the same (or greater)
capacity, with the extra two used solely for backups.
The next time somebody shows you their monster size drive or array, ask
them where do they keep the other two? <g>
Scott.
