ADSM-L

Re: Is network backup still a good idea?

1998-10-21 16:01:24
Subject: Re: Is network backup still a good idea?
From: Laszlo Nemeth <laz AT SYBASE DOT COM>
Date: Wed, 21 Oct 1998 14:01:24 -0600
Yes. every couple of years disk REALLY outpaces tape. then something
comes along in tape/disk that fixes the problem. it used to be no exclude
lists.
or disperce data. disperce data was fixed with nfs servers/raid. exlude
lists
next generation will be a diff/rcs type of thing on the data ie:
doc vesion1
doc version2
doc verrsion3
only the added 'r's and version would be backed up

dunno if adsm will do this in the future

from the technical side you can almost always get around the problem, its
just a money
issue. here we just ask for more money or tell management we will turn off
backups
to lowest priority machines moving upwards untill our system is full. since
laptops
are lowest priority and managment all has laptops we get the money ;-).
same applies
to a university. when i worked at cu we charged per machine/amount of data
to back it
up.


laz




Roger Deschner <U52983 AT UICVM.UIC DOT EDU> on 10/21/98 11:46:59 AM


ADSM is truly best-of-breed at what it does, but we are beginning to
question the whole notion of network backup of desktop systems.

Why? Because they are getting so huge!

I can go to Best Buy or a mailorder place and get an 8gb disk drive for
under $200. Then I can go out with my camera and take a bunch of pictures
of trees changing colors in the mountains, microbes under a microscope,
or whatever, mess with them in Photoshop, or I can reprocess the census
ad infinitum in SPSS, and fill it up. And then ADSM comes along and backs
that up.

Even with: 100mb switched Ethernet and DLT tapes, a data explosion is
underway. Disk drive technology is outpacing network technology,
software, server technology, and tape technology.

When we started in ADSM, we assumed it would be just a small thing that
we could move around to different servers. Now, it is taking a
significant part of a very large system, and we can't move it without
great effort even if we wanted to.

What are we backing up? Parkinson's Law applies to disk space. Data will
expand to fill the available disk space. Exclude files that skip over
duplicate copies of the operating system itself are but a drop in the
bucket.

In this brave new world of snowballing digital junk accumulation, ADSM
does not scale too well. Oh, it does its job just fine. But as we move to
this brave new world where a single desktop system may contain hundreds
of thousands or even millions of files, the ADSM database begins to get
in the way. Although it is a very efficient thing, and its
commit/rollback architecture makes it extremely reliable, it is just
keeping too much information about its contents, in an environment where
the number of stored objects is exploding by orders of magnitude.

And the objects themselves are growing too! I'd love to be in the data
tape business right now.

We can probably push ADSM farther into this brave new world than most of
its competitors, because of its basically superior design - but even it
won't last long. It's an architectural problem, not a technical one; ADSM
is technically superior on

We could buy a cheap, high-capacity tape drive for each desktop unit, and
let each owner (NOT!) back it up. This puts us back where we started when
WDSF (ADSM's predecessor) was invented - valuable corporate data at risk
because some doughhead didn't understand the importance of backing up his
own PC. ADSM has directly saved more than one doctoral dissertation and
funded research project here at UIC. I don't like going back to
individual workstation backup, but that's where we're headed, as the
fundamental equations of backup change, again.

I'm fishing for a general direction, and I have not mentioned any
specific platforms or technology except to illustrate the basic problem.
This is a cross-platform industry problem. Comments? Debate? No flames -
I am asking this as an open ended question, for which I'm afraid there
is no "right answer". It's a whole new world, again.

Roger Deschner      University of Illinois at Chicago     rogerd AT uic DOT edu
Aliases:             USUICZ3P at IBMMAIL            u52983 AT uicvm.uic DOT edu
== All of our systems will be fully Year 2000 Compliant by Jan 1, 1900 ==