[Discuss] Backing up LVM partitions using snapshots

[invalid at pizzashack.org (Derek Martin)]

On Fri, Dec 16, 2011 at 01:48:52PM -0500, Richard Pieri wrote:
> On 12/16/2011 12:38 PM, Derek Martin wrote:
> >These numbers are false though.  If you need to back up 300TB of data
> >using tape, you need way, way more than 300TB of tape.
> 
> Even if you need three times as much tape capacity as disk capacity
> then it still costs an order of magnitude less to buy the tapes than
> it does to buy a new storage frame.

True, and tape drives are free.

> >tapes (3TB uncompressed) run about $110, a 3TB drive runs about $130.
> >So how much are you /really/ saving by using tape?  My guess is, over
> >time, even with the expensive EMC solution, the savings for tape are
> >negative; substantially more so with cheaper disk solutions.
> 
> This is marketing again.  Your $130 cost for the 3TB disk ignores
> the cost of the chassis to house it and the cost of the power to run
> it. Even a "cheap" chassis is a 5 digit expense.  

Likewise the $110 cost of the same amount of tapes ignores the cost of
transport & off-site storage, on-site storage, and the actual tape
hardware.  You'll either need a monstrous robotic library to manage
this much tape, or you'll need to pay the salary of a human to manage
it all for you full-time, or more likely, a little of both, since the
tapes need to get into the robot *somehow*, and the human will screw
it up without automated help...  Neither of those are cheap.  You're
talking at least 100 tapes per full back-up (1.5TB uncompressed, 3.0TB
compressed per tape), x2 for off-site copies, x your retention policy
factor, plus whatever tapes you need to deal with your incremental
back-up scheme.  That's a metric arse-load of tapes, up to half of
which need to be transported to your DR site with some frequency.  

By contrast, The disk solution can be very nearly fully automatic, once
it's set up, since there's no media to swap.  There's no transport
needed, as once the back-up is complete, the primary can replicate to
the secondary fully automatically over the ether.  All you need is
periodic maintenance of the disks that fail, and some amount of
attention to making sure it's still running smoothly.

> I can expand tape capacity by that much for the same cost, with
> redundancy, and I can infinitely expand my tape capacity without
> buying more hardware to house and power it.

...because tapes and tape drives occupy no physical space, and tapes
are magically self-inserted into the tape drive?  Or perhaps you keep
them in an altnernate plane of existence, operated by data archive
seriphs out of the goodness of their souls, connected to your servers
by MPFL (metaphysical fibre link)...  Also how does the transfer rate
of 300TB to high-speed tape compare with 300TB to disk?  It doesn't
matter if you can do it, if the back-ups take longer to run than it
would for a human to type in the data by hand...

Everything has a cost.

> So yes, I really am saving money with tape even with redundant
> media.

If you say so, but I seriously doubt that.  But I don't think Mark is
talking about backing up anywhere near 300TB, so let's make this a bit
more realistic:  You have one terabyte to back up.  For tape, you
need:

 - Two 3TB LTO tape drives (to make your redundant copies): $1500
 - four tapes for full backups (you need at least two sets, so you're
   not writing over your good back-up, x2 for redundancy): $200
 - At least two tapes for incremental back-ups, probably more: $100
 - probably a computer to connect this to, to manage it all: $500
 - At least 6 more tapes, every 6 months, probably more: $300

Tape solution: $2600 for the first year, $600 per year after that,
ignoring hardware maintenance.

For the disk solution, you need:

 - 2 1TB drives:  $250 (certainly less, but close enough)
 - a computer to connect them to, $500 for adequate PC hardware (or
   less) x 2: $1000

Disk solution: $1250 for the first year, less than $250 per year
after, ignoring non-disk hardware maintenance.

Prices are from pricewatch, the cheapest 3TB LTO5 drive I saw was
$763.  1TB disks are available for under $100, I allowed $125 per
disk.  Other PC hardware was estimated, you can build a cheap, capable
PC for less than $500.  Of course with either solution, you might also
already have a machine that can serve the function, eliminating that
cost from the equation.  This also doesn't include the time it takes
to manage: the tape drive solution requires a human to be on-site to
swap out tapes; the disk drive solution requires very little human
labor.  Both solutions should involve some amount of management to
make sure that back-ups happen, and to verify they are usable, so I
assume that cancels out.

The disk solution wins in every possible way.  It's faster, more
reliable, and costs less than half as much.  Plus you can have 3
copies of your data *accessible* on line simultaneously at all times.
You can afford to make this a cheap RAID solution both on- and
off-site and still come out ahead.  Note that you also still come out
ahead with disk if you drop the need for off-site redundancy.


-- 
Derek D. Martin    http://www.pizzashack.org/   GPG Key ID: 0xDFBEAD02
-=-=-=-=-
This message is posted from an invalid address.  Replying to it will result in
undeliverable mail due to spam prevention.  Sorry for the inconvenience.