In talking with a whole ton of people about this, this is the tentative
idea:
This is for document retention, so the recovery of a whole server isn't a
requirement. If I can get back a MSSQL database file, I can restore it
anywhere. Same goes for a .doc or .xls file - I don't need the whole server.
So my plan is to create a storage pool, call it "ARCHIVE_STGPOOL" or
something.
Load a bunch of scratch tapes in the 3583, run my monthly archives to that
stgpool. When they're done, check them out, mark them offsite, etc. But we'd
run incrementals, not monthly fulls, to keep down the db and tape usage.
We're a fairly small shop (now) and only use around 3 tapes a day.
I *love* that TSM is so flexible and I hate that I get to agonize and debate
any changes to it for that very reason. We've been bitten really hard in the
past due to poor planning, so I'm trying to plan for my next 4-7 years now.
:-)
-----Original Message-----
From: ADSM: Dist Stor Manager [mailto:ADSM-L AT VM.MARIST DOT EDU] On Behalf Of
asr AT UFL DOT EDU
Sent: Friday, July 16, 2004 10:08 AM
To: ADSM-L AT VM.MARIST DOT EDU
Subject: Re: Thoughts on Monthly Archives
==> In article <s0f7c1a7.077 AT health-es2.health.qld.gov DOT au>, Steve Harris
<Steve_Harris AT HEALTH.QLD.GOV DOT AU> writes:
> at 1% , 1-(0.99**30), or about .25
> at 2%, 1-(0.98**30) , or about .45
> at 3%, 1-(0.97**30), or about .60 (Please feel free to correct my maths
if I'm wrong - probability was never my strong point)
I think your math is good; I'd add though that there's a -GREAT- deal of
locality of reference: Or in other words, if 3% of your files change a day,
then for user filespace probably 95% of the files that change tomorrow will
be the files that chaged yesterday, and so on. I think that 25 - 30% for
userdir space is probably about right.
> Now, of course its often the same files which change day after day, so
> real experience should be better than this, but at the time, I decided
> that the overhead of mainitianing two TSMs (and two clients per node)
> wasn't worth the benefit, and went with archives.
But I would disagree with your logic;
In place of the incremental possibilities, which would have led you at worst
above to backing up 60% every month, you're choosing to back up 100% every
month, without fail.
I think that this represents a substantially more costly strategy.
In fact, given the monthly-for-five-years figure and your numbers above,
it's about twice as expensive in facilities to archive monthly than to run
incrementals with similar retention characteristics; If my guess is closer
to correct, it's three- to four- times the cost.
For a small amount of data, this may be cheaper than the human
organizational time to build two sets of nodes; but by the time you get even
medium size, I think it would be pretty expensive.
- Allen S. Rout
|
