Reply to MESSAGE FROM: MARIST

>        What I am looking for are examples of ADSM specific hardware
>implementations.  For example, what hardware are you using for the server
>and tape drives.  How you have configured your media transport, including
>topology, if dedicated specifcally for ADSM only and if you pass data across
>routers and bridges.
>
>        Currently we are using our exsisting network to pass data from the
>client(Netware) to the server(IBM9121 MVS) using TCPIP.  The data passes
>over ethernet thru a Cisco router, to FDDI to another Cisco router, to
>ethernet into a IBM 8232.  In this configuration it take at least 10 hours
>to do a full server restore, about 4 gigs of data.  We are in the process of
>setting up a dedicate token ring from the clients to a IBM3172.  Is there a
>better solution?  I read about  200 client setups,  but our initial 4 client
>setup is a non workable solutions.
>
>Server version:  v1.r2.level 8 connection states 1.1.13
>Client version:    1.02r NLM
>
>

*** Comments from STEWAJM - Stewart, Mike; 02/15/96 07:31am:
With the following configuration,
I get about the same throughput you've experienced, about 5 MB/min:

90mhz Pentium/Windows/Microsoft TCP/IBM TR
   |
   |  -> TR utilization < 10%
   |
 3172-3   -> 3172 utilization < 20%
   |
 9672-E03 -> machine utilization during tests about 10%
             ADSM uses about 30% of 1 processor
             No paging.
            DASD/Channel Util < 20%


Throughput measured is:
ADSM client "megs backed up"/time (compression turned off).

Adjusting ADSM DSM.OPT TCPBuffer/window sizes have no affect.

Starting a backup on a second machine on the ring does not
have a significant impact on the throughput (ie, both machines
get about 5 MB/min.)

My guess is that the bottleneck is the client machines.  Maybe
the ISA Ibm token ring card?