2+ TB storage options for Linux/SCSI computer

[qzectb]

I have some data sets that add up to around 2 TB, and I need a
convenient and inexpensive way to archive and access them from my older
(vintage 2001) Linux PCs (w/ internal SCSI and IDE controllers) without
an inordinate amount of human intervention.

I know that a RAID is one option, but are there others? In the old
days, I used to keep my large data sets on Exabyte 8 mm tapes (5 GB
capacity each) and then on more expensive Mammoth-2 tapes (~50 GB each,
$70 or so per cartridge). But 1 TB equals 200 of the 8 mm tapes and
20+ of the Mammoth-2 tapes, and that is neither convenient nor terribly
cost-effective. Also, I had a lot of trouble with read/write errors and
the need to clean the drive heads very frequently.

Are there more modern tape storage systems with higher capacity, higher
reliability and/or lower cost? It would have to be a tape device that
is seekable (e.g., using Linux mt commands), not just something that
streams from start to end and changes direction extremely reluctantly.
And I don't think I'd want to deal with anything under 100 GB per tape
cartridge (if such a thing exists), because of the manual labor of
switching tapes (we've had bad luck with tape libraries).

What about optical storage? Should I consider some kind of DVD
jukebox? I have no experience with this medium, so I don't know what's
out there or what it costs.

If I go with a RAID, which makes/models should I consider? Read/write
speed is not nearly as important as low cost per unit storage.
Reliability isn't a huge issue, because the data sets are replaceable
(albeit with some inconvenience) if something fails. Scalability would
be nice, though it would probably take time for me to outgrow a 2 TB
RAID.

Thanks for any help .. I used to be up to speed on storage in, oh,
about 1995, but it's hard to stay caught up since it's not in my job
description.

 

[qzectb]

I forgot to mention that the data sets need only be written once, so I
don't need a storage medium that lends itself to frequent updates.
DVD-R would work, for example.

 

[Ken Maltby]

I forgot to mention that the data sets need only be written once, so I
don't need a storage medium that lends itself to frequent updates.
DVD-R would work, for example.

The burned die DVD disks have no proven reliability for
long term storage. Digital Tape has reliable proven storage
parameters. The modern hard drive has well established
and practical storage parameters. Five 300BG drives would
give you 1.5 Terabytes, 3 Terabytes for a 10 pack. At $0.50
a GB you would be talking $1000, for your 2T.

Luck;
Ken

 

[Paul]

I forgot to mention that the data sets need only be written once, so I
don't need a storage medium that lends itself to frequent updates.
DVD-R would work, for example.

250GB hard drives can be had for $113. At that price point, they
can be either SATA or IDE. Eight of those gives you 2TB.
Doing software RAID5 would require at least one more drive
(and protects you against a single drive failure). Using one
disk to protect eight disks, is pretty economical (assuming the
software can handle that many disks).

Then you needs some cheap non-RAID controllers, and the software
does the RAID part. So maybe $1200 gives you a backup solution
with random access characteristics. A little compression in
the backup method could save you a few bucks as well (fewer
disks if the data can be compressed).

While DVD-R sounds enticing, I wonder if you could get archival
quality burns done completely hands free. It is pretty
annoying to find a DVD that reads in one drive, and not
another, and you would not want to find that kind of
behavior when you least expected it. And the RAID array
would read and write a bit faster than the DVD would.

As for packaging the disks, I would want a machine you
could shut off and unplug when not in use (unplug both
power and network connection). That helps eliminate a
ligntning strike as a possible cause of data loss. You
could even pop the machine in the car and store it
offsite. While trays with a lock and key can be used
to make removal and transport of individual
drives easy, nine trays at $30 each adds $270 to the
price. Otherwise, I'm assuming you have an old computer
case with enough room to store the hardware. You
might need to pick up a beefier power supply.

While 500GB drives are shipping, they are not the
most cost effective storage devices right now.

Paul

 

[qzectb]

Thanks to all for the extremely helpful advice.

Here's one more option I just learned about that sounds especially
appealing to me: network attached storage (NAS), such as the Infrant
ReadyNAS 600 - 1.0 TeraByte (4 x 250GB SATA HDD). Cost is around
$1,200.

What I like about what I've read is that you don't have to have a
dedicated server computer for it; you just hang it off the LAN and NFS
mount it on your LInux computers. And unless I'm mistaken, it's
infinitely scalable in the sense that if you need another TB, you just
hang another one on your net.

The main drawback I'm aware of is that there is no computer that has
local access, so reading/writing to the NAS always ties up the
bandwidth of your LAN. But that would be the case anyway for any of my
computers other than the one a RAID was physically connected to.

Is anyone aware of any other reasons to consider a traditional RAID
over a NAS?

thanks

 

[kony]

Thanks to all for the extremely helpful advice.

Here's one more option I just learned about that sounds especially
appealing to me: network attached storage (NAS), such as the Infrant
ReadyNAS 600 - 1.0 TeraByte (4 x 250GB SATA HDD). Cost is around
$1,200.

What I like about what I've read is that you don't have to have a
dedicated server computer for it; you just hang it off the LAN and NFS
mount it on your LInux computers. And unless I'm mistaken, it's
infinitely scalable in the sense that if you need another TB, you just
hang another one on your net.

The main drawback I'm aware of is that there is no computer that has
local access, so reading/writing to the NAS always ties up the
bandwidth of your LAN. But that would be the case anyway for any of my
computers other than the one a RAID was physically connected to.

Is anyone aware of any other reasons to consider a traditional RAID
over a NAS?

thanks

Personally I prefer "both", a regular system as the NAS,
using raid, over gigabit ethernet. Main drawback- takes up
an extra cubic foot of floor space. You can get 250GB HDDs
for about $90 ea. now, less after rebates. Figure $120 or
so for a refurb'd system from someplace online if you can't
source something suitable locally (or already have an older
system to devote to it), throw in an SATA card and a GB
network adapter and total cost for that 1TB was around $520,
and if you choose a case with enough room for more drives,
the next TB is a little under $400 more, even less if you
factor in that HDD capacity per $ keeps rising.