[HH] AoE controller

[Bill Bogstad]

On Thu, Jul 26, 2012 at 4:00 PM, Tom Metro <tmetro+hhacking at gmail.com> wrote:
> Bill Bogstad wrote:
>> What are your use cases?
>
> One scenario I have in mind is a large pool of non-RAID storage used
> strictly for backup and "near-line" storage.

Ahh,  another person whose noticed that the cost of disks vs. the cost
of high capacity tapes are getting
closer and closer together.   Still not quite there though:

http://www.newegg.com/Product/Product.aspx?Item=N82E16840995035

1.5TByte tape (uncompressed capacity) for around $50.   My low end
rule of thumb at the moment for raw disk pricing is $50 per TB so tape
is cheaper.   OTOH, the tape drives to access those tapes are still
moderately expensive:

http://www.newegg.com/Product/Product.aspx?Item=N82E16840121082
$1862 (reduced from $3200)

The answer in the tape market has always been to use a jukebox.
I've always thought you could do the same thing with disks without
having to incur the high cost of the drives to read them.   However,
I've yet to see anyone go down that path   It would seem to me that
given that SATA is hot-plug that the only
part that wouldn't be off the shelf technology would be the carriers
for the drives and the pick & place mechanism.

The advantage of SATA vs. tape is that adding extra "active" vs.
"dead" slots in a disk jukebox is going to be cheap for SATA while
expensive (~$2K each) for tape.   If you used some kind of backplane
mechanism for your disk jukebox, you could vary the number of active
vs. dead slots fairly easily depending on how many live spindles you
wanted in any particular jukebox.  Disks already have readable serial
numbers so no extra machine readable labels would be required to keep
track of things, if someone put a disk in the wrong slot.  With a tape
based jukebox, when the higher capacity drive/media combination comes
out, you not only have to buy new media, but you also have to replace
the tape drives.   With a SATA jukebox, higher capacity media is
backwards compatible at the interface level with the jukebox so you
only have to purchase new disks and could mix & match any number of
media capacities in the same jukebox.

Active slots are cheap enough with SATA, that you might even be able
to do away with the jukebox mechanism entirely on something used
solely for off-line storage/backup.   Once a day/week someone could
manually pick & place the next 10-20 disks into the active slots as
needed for your next round of backups.   Are you really backing up
more then 20-40 TB a day/week?

I suppose reliability (both short & long term) of the media might be
of concern with a disk based system.
Not sure if the best way to deal with that would be multiple copies,
forward error correction, or a secondary system that slowly migrated
disk images onto tape using a single tape drive for long term
archiving purposes.   Maybe a combination of all three.   Not sure of
the cost implications of this.

Admittedly my ideas are pretty much an attempt to mimic the
functionality of traditional high-capacity tape systems using disks.
 Without the automatic pick & place mechanism, it wouldn't be suitable
for near-line systems.  Still, I'm surprised that I haven't heard of
anybody doing something like this cheaply.   Maybe it is just that
people who have these kind of backups requirements are conservative
about the technologies that they use.    Disks are "unreliable" so we
can't base our off-line backups/archives on them.   Or maybe the price
advantages just aren't there yet.

Bill Bogstad