OK, we are just starting to play around with multiple network paths
BECAUSE up to now we backed up our 2 TB data base over a single FDDI
interface to an adsm server on the same subnet and the network was NOT the
bottle neck
We would run 30 concurrent sessions on the client machine (to the adsm
server) and because we were doing client compression our limit was the
client CPU
I think we are at the max number of CPU's that may be placed in this
specific client box one of these big Sun E 10K's and still really the single
FDDI isn't a problem.
Now if we didn't compress the data the network would be a problem BUT we can
compress the data and send the compressed data in less time than we could
send the uncompressed data (with unlimited band width)

Yep, ok, so we dump about 600 GB of compressed data into a diskpool nightly,
we first go to disk so we may have 30 concurrent client sessions (and only 4
3590 tape drives)  THEN once the client is done (implies our diskpool is
large enough to hold one 24 hour period of backups/archives) we migrate
it... one, single, migration process has to move that 600 GB from disk to
tape.   Now this adsmserver's clients have kind'a out grown our design
because it takes about 20-ish hours to get the data to the adsm server and
about 20-ish hours to migrate it to tape but that is 600 GB from a single
client (and we do have other clients on this adsm server)
and remember all that 600 GB is already compressed, that is 2 TB client
defined file space and the .dbf files that occupy it.

If one has client nodes that are more alike the processes work out better,
keeping more tape drives busy and getting the data off the disk and onto
tape sooner.

We have 6 SAP production instances (and the TONS of CN, QA, PF, etc...
instances) being backed up by adsm and we are really happy with it.

later,
       Dwight

Dwight Cook
BP-Amoco



> ----------
> From:         Michael Klatt[SMTP:Michael_Klatt AT NMC DOT DE]
> Reply To:     ADSM: Dist Stor Manager
> Sent:         Thursday, April 20, 2000 10:01 AM
> To:   ADSM-L AT VM.MARIST DOT EDU
> Subject:      more: Question regarding to the number of drives
>
> Hi all,
>
> the reason to talk about it is, that I have had a discussion with an other
> vendor of a backup product about performance in an SAP enviroment. My
> questionsTSM related  which are still not answered yet are as follows and
> maybe
> someone can help.
>
> -    the  SAP backup description  talks about multiple sessions. These
> sessions
> execute in parallel. Does this mean different ways (different Lan
> connections) to the *SM Server? If yes, how is it managed? Can I use a
> switch in
> between? For example 4 times 100Mb out of the node into the switch as
> an
> uplink to 1 Gb link to the server?
>
> -    a storage device (tape drive) must be associated with a session.
> Isn't it
> possible to write first to a diskpool and migrate then into a library
> using the
>      number of drives available in that  library? If yes, will the max.
> number
> of drives equal in the number of sessions or is it like Gary explained
> "one
> single    migration process per unique node's data in the diskpool"
>
> -    are there any informations available to get a better knowledge about
> backup
> of SAP date by using  TSM .
>
>
> many thanks in advance
>
> Michael Klatt
>