ADSM-L

Re: "saving" tapes (or management's ideas...)

1999-01-25 17:02:17
Subject: Re: "saving" tapes (or management's ideas...)
From: Dwight Cook <decook AT AMOCO DOT COM>
Date: Mon, 25 Jan 1999 16:02:17 -0600
     *** numerous include/exclude statements...
        each file on the client is run through (bottom to top) the
     include/exclude list until a match is found.  If your hit rate of a
     client file in your include exclude list made a nice bell curve around
     the median entry (which it probably won't since these lists tend to
     EXCLUDE files and any comparisons that purculate out the top are
     included by default) you would have say :
     10 million client files
     10 thousand excludes
     nice 50% average hit rate
     yields 50,000,000,000 (50 billion comparisons)
     BUT like I said, mainly excludes with default includes if no match
     found so I'd use 90% for the average excludes compared prior to a
     decision being made so we would have 90 billion comparisons.
     OK, how fast is your CPU ? ? ?
     Still boils down to 0 is way less than 50 billion which is less than
     90 billion.

     That's why there is the "DOMAIN" statements... domains are file
     systems / volumes / (platform specific whatever) and if the fs/vol/xxx
     isn't in the domain, the files within never make it to the
     include/exclude's.  100 files compared in an incl/excl of 101 with the
     top entry being a wildcarded entry to catch all 100 of your compared
     files will still yield 10,000 comparisons just to back up ZERO of the
     files... if the files were all on a single volume and that volume
     wasn't listed in the domain statement... 1 compare would exclude all
     of them (more or less... they wouldn't even be built into the list to
     run through the inclexcl)

     this still leaves you with a ton of manual work to seperate & isolate
     data on your client servers...

     look at it as job security !

     Hope the adsm info is a little help
     later,
           Dwight




______________________________ Reply Separator _________________________________
Subject: "saving" tapes (or management's ideas...)
Author:  zvika (zvika AT AESERV.TECHNION.AC DOT IL) at unix,mime
Date:    1/25/99 11:20 AM


I am currently involved in the painful process of trying to convince
the "powers to be" to spend $$$ for "backup h/w"  and I need some advice.

We are running ADSM on an AIX machine to backup the computers in our
department (we are a technical university). We have a 3570-B12 (that's
2 drives and 19 cartridges), and that is almost full. We don't have a
copy pool, and don't archive (well - got 1cartridge dedicated for that
just so I could get the feeling of how it works, but it's not in production
for the users).

We had a policy of 60 days + 2 versions is file exists and 90 days/1 version
for deleted files. The only action approved to me when we choked on space
was to lower these values to 30/2 vers. (exists) and 60/1 (deleted), and
we more or less saved the space of 1 cart (around 10.5GB).

I requested to get a 3575, make it our primary backup pool and create
a copy pool on it, and dedicate the 3570 to an acrhive pool + db backup.

The enthousiasm was very small to my request, and I was asked to check the
option of minimizing the files on the backup media, by having (some of) them
on other/cheaper media, such as on CD-Rs, sort of HSM from tape to other
backup media. As far as I know there's no such possibility with ADSM,
and I was therefore asked to check for the possibility that files untouched
for 1 year (for example) would be deleted from backup and sent to some
archive. Is this possible under ADSM ?
I guess I could do a find old_untouched_files >> my_adsm_exclud file,
and then the files will be expired, and I could archive them....

A few questions about that (mad, IMHO) idea:

1. What will  the expected performance penalty be when performing incrementals
   with exclude lists of tens of thousand files (upto, around 150,000 on
   my major client, the NFS server which is also the ADSM server) ?

2. Any "simple" way to make sure, that if a file's changed, its status will
   be changed from "no backup, once archived" back to "normal" backup ?

3. Do you know of any very cheap storage solutions (fully automated, we can't
   do it manually for lack of manpower) that is available to work with ADSM
   (cheap = relatively to 3570/3575 solution).

4. Any ideas on how to approach the problem ? Other solutions ?

I would appreciate your input.


Regards,
/Zvika



~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
I cherish the environment - this message was composed with 100% recycled bits.

Zvika Bar-Deroma
Systems and Network manager                       Phone: (+972)-4-829-2706
Faculty of Aerospace Engineering,                 Fax  : (+972)-4-823-1848
Technion                                          Home phone:
Haifa 32000                                              (+972)-4-823-5562
Israel

Internet      :   zvika AT aeserv.technion.ac DOT il
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
<Prev in Thread] Current Thread [Next in Thread>