If you really expect 542Gb of incremental data per night then your diskpool
should ideally be at least equal to this amount of data. This should stop
migration from kicking in and help your performance.

Take a look at the network also. I don't know what your topology looks like,
but it's very easy to overlook the fact that having your servers on
different subnets from the ADSM server places a substantial load on the
network routers. It's easy to choke these routers when running multiple
backups simultaneously.

You can introduce additional routers to ease congestion, but then you have
to remember to update the static route tables on you clients to the ADSM
Server and load balance accordingly.

Alternatively you could look at multi-homing some of those servers on
different subnets, placing an interface on the same subnet as the ADSM
Server.

Nathan

        -----Original Message-----
        From:   Steven Thornhill [SMTP:steve_thornhill AT UK.IBM DOT COM]
        Sent:   Thursday, December 16, 1999 9:45 AM
        To:     ADSM-L AT VM.MARIST DOT EDU
        Subject:        Re: ADSM system scaling

        Personal opinion is to double up on the number of ADSM servers.
Split the tape
        drives across two machines will increase the data throughput. Also
this will
        give you a little more resiliance.

        Steve Thornhill - IBM ETC
        Int       - 255104
        Ext      - 01705 565104
        Mob   - 0378 289770
        Email - steve_thornhill AT uk.ibm DOT com


        "Patten, Stacy H." <shpatten AT SAINTFRANCIS DOT COM> on 16/12/99 
15:18:22

        Please respond to "ADSM: Dist Stor Manager" <ADSM-L AT VM.MARIST DOT 
EDU>

        To:   ADSM-L AT VM.MARIST DOT EDU
        cc:    (bcc: Steve Thornhill/UK/IBM)
        Subject:  ADSM system scaling




        Having never scaled an ADSM setup I would appreciate input for our
        situation:

        Current Hardware:
                1 - H50  -512MB / 1-300MHz
                24GB disk pool
                1 - 3494 Library w/ 3 3590 drives
                1 - ATM NIC (emulating token ring)
                1 - 100BT NIC (not currently in use)

        Current Backup:
                148GB

        We expect that by Dec 2000 we will be at 542GB a night.  We barely
make the
        current backup routines, and this number will increase to 223GB with
the
        addition of Exchange, SQL, and ORACLE clients.  Our backup window
10:00pm to
        6am so we should have plenty of time to perform this, with the
exception of
        migration consuming our tape drives twice a night.  Our servers are
on
        different subnets, so the backup traffic goes across our ATM
backbone and
        back before it hits the ADSM server.  Also, most of our servers are
running
        100BT ether with some on 16MB token ring.  We have (over?)estimated
the
        following hardware setup to hopefully solve our capacity issues
until at
        least the end of 2000.

        Proposed Hardware:
                1 - H50 - 1GB / 2-300MHz
                150GB disk pool
                1 - 3494 Library w/ 8 3590 drives
                2 - ATM NIC (emulating token ring & ethernet)

        Is this enough machine?
        Is this enough disk pool?
        Is this enough tape drive?
        Is this enough bandwidth?



        Stacy Patten

        Senior Network Engineer
        St. Francis Hospital
        phone- 918-494-9465
        pager- 918-646-8497
        email- shpatten AT saintfrancis DOT com