Al said:
I was talking about the media, and how long it will be readable.
backup, archive. what's the diff ?
The diff is that you do a backup every day and if you're doing it wisely
you're maintaining at least several days, and possibly several weeks, but
the purpose is disaster recovery. If the media fails after 30 years it
doesn't matter because it's going to get overwritten every couple of weeks.
On the other hand, backup has to be fast and simple or it doesn't get done
anywhere except in large organizations that can put someone in charge of it
and someone else in charge of watching the person in charge to make sure he
does it.
Archiving gets done when there's data that needs to be stored for a long
period of time but does not necessarily have to be immediately available.
It's usually relatively infrequent occurrance but the purpose is to store
and preserve the data, which means that the media should be durable and
there should be multiple copies kept because _no_ media can be considered
completely reliable.
My archive frequently consists of
taking a tape out of the backup cycle and putting it in an offsite
vault, for posterity.
Which if the data is valuable is asking the universe to kick you. Unless
you maintain two or more copies.
At one point I had 18 years of tapes. QIC, 8MM,
DDS, and DLT, from just about every OS you could name. We knew that
getting data from the QIC cartridge from the old MASSCOMP would be a
PITA to read, but know that we could get a service to do it if
necessary.
If the data on the tape is readable. Personally, I have data on my server
going back to 1988. I don't see any need to periodically copy that to the
tape dujour and throw it in "the vault" and then worry about resurrecting
some antiquated tape format later. It all gets protected by my regular
backup. When I fill up the server I just expand it.
This kind of archive made sense for the business we were
in. From time to time I took the oldest tape I still had equipment for
and did a readback. I have to say that the reliability was very very
good.
Given the nature of the company I worked for we occasionally got
involved in software IP casses to prove prior art. Once We got 800bpi
tapes from a major university written by a UnixV6 system. We could
get a service to read the bits and FTP us the data, but the format of
the backup data bore no relationship to any current software. One of
the guy I worked with was a unix file system implementor in the V6
timeframe. I watched hin find the data blocks (from the CD, with grep)
that were the C code for the file system drivers. He them hacked the
read-only part of that code to serially read the rest of the of the
blocks and reconstruct the file system. We recovered docs and C code
that proved that want some graphics tpy that a comapany was trying to
patent was being done in a university in the 70's.
Ain't open source great.
Nope, having someone who knew the formats was great. It may have escaped
your notice but Unix has _never_ been open source. Linux and the later
versions of BSD, but neither of them is based on AT&T code anymore. If you
needed to see the source to V6 in order to do what you needed to do, you
might find someone with a source license but if you didn't you'd likely end
up paying AT&T or Novell or SCO or whoever had it at the time about as much
for the source as you would Microsoft for a source license to XP. And that
assumes that they even _have_ the source anymore.
The first mag tape I was involved with was 200BPI 7-track, (upgraded
to 556 right after I started working). I also worked on data
retention policies and technology in a major bank.
Well that's nice. But what does it have to do with the economics of backup?