[conspire] DVD as backup media

Rick Moen rick at linuxmafia.com
Mon Aug 11 02:15:54 PDT 2003

Quoting Greg Dougherty (gregd at molecularsoftware.com):

> I paid less than $750 for my whole computer (not counting the monitor,
> which I already had).
> I like backups.  I'm not going to double the cost of my system so that
> I can make a reasonable number of backups.

I'm certainly not saying that's not an unreasonable stance.  It's one
many people take (sometimes consciously, sometimes not).  

But one vital question hasn't been addressed:  What's your _data_ worth?
The answer to that question basically tends to determine how seriously
you (thereupon) take backup planning.  How seriously you take backup in
turn determines what strategies you take.

Many people don't know the worth of their data until they've lost a few
months worth of hard (and possibly irreplaceable) work product.  (Of
course, if you're lucky, you have little or nothing on your machines
that poses that sort of problem.)

There was a manager once at a firm I worked at, fifteen years back, the
VP of Engineering at a software company.  I was the grunt who pretty
much ran company IT (which we then called "MIS").  He wanted me to spec
and buy a NetWare server, which was going to do basically everything
that department needed -- IPX/SPX (Novell) file & print serving, NFS,
lpr printing, version control, software building, Sybase, Oracle, and
backup.  Budget was $20k.  I looked at the list of functions, balked,
politely disagreed with his assessment that these should all be done
by one machine, and handed him a counterproposal of several machines
that would split up those functions -- collectively coming in under the
same budget.

He considered me out of line, complained to my manager, and outsourced
the purchase to a contractor -- who in due course delivered the
requested box, and deployed it.  Six months passed, seven months....  It
gradually dawned on everyone that the new server was progressively but
slowly corrupting all data every time it passed through RAM (even though
the RAM tested OK, and the syndrome persisted after swapping out the RAM
and other parts).  

The VP of Engineering was now faced with every software engineer's worst
nightmare:  He could have us restore any backup set from prior six/seven 
months that he wished.  Did he want an old backup set?  That would
minimise the corrupt data, but lose 100% of what every engineer had been
doing for the last half-year.  Or he could restore a recent backup and
regain some of the company's recent history, except that more files
would have mush in them, the closer he got to the present.

In the end, the firm decided to discard seven months of everyone's work --
and fire the VP of Engineering.  (I did my best to keep a low profile,
and say nothing sounding even remotely like "I told you so.")

One lesson of that story is that backup _isn't_ protection against all
threat models against one's data.  But another is that the main value of
computers tends to _be_ that of the data itself -- in that case, six
months of a company's work and a VP's job.

> Tape drives haven't kept up with hard drives.  It sucks, but it's true.

I think you're right.  But losing a huge amount of work data can also suck.
Especially when a cheap, used DDS2 drive from Action Computer ($200?)
and a small pile of $6 tapes would have given you the means to recover.

On the other hand, some people do just great with more-casual and/or
cheaper backup methods, or just don't keep much data they'd miss on-disk.  
Whatever works for people is good.

My job, working for companies, usually involves finding and considering
realistic threat models they hadn't fully considered, and trying to plan
for them appropriately.  Usually, when I put it in terms of risk
assessment and control, good tape backup (applied the right way) ends up
being considered cheap insurance.

Cheers,                            Ceterum censeo, Caldera delenda est.
Rick Moen
rick at linuxmafia.com  

More information about the conspire mailing list