James Riehl: if you see this and want me to open a call  or can discuss
offline my notes id is
dlancast AT au1.ibm DOT com

I've got real bad performance on a Solaris system when FTP performance is
fine

I have:

Server: TSM 4.1.2  Solaris 2.8

ndd  tcp_xmit_buf   64K
       tcp_recv_buf    64K
       tcp_cwnd_max 1024K

DSMSERV.OPT
      TCPWindowSize  0

Disk pools in this case are on a T3 Fibre Array, I can't fault it


Client: TSM 4.1.2 (IP22150)  Solaris 2.8
ndd  tcp_xmit_buf   64K
       tcp_recv_buf    64K
       tcp_cwnd_max 1024K

DSM.SYS
       TCPWindowSize  1024
       TXNBYTESIZE 8192
       TCPBufSize   256

Two client systems

1) 100MB interface
       TCPNODELAY NO in DSM.SYS
 I get 13GB in 30 mins using BACKINT (TDP R/3 for Oracle) , using 3 sessions
and reading 4 files per session.
I can't drive it any faster, the performance seems fine for 100MB hme
interface
NB: When TCPNODELAY was set to YES, backup time was 60 mins

2) 1GB interface (ge0), back to back to the TSM Server (no switch)
FTP performace 20GB in 30 mins OK
BACKINT (Same params as 100MB)  200GB in 17 Hours  (Ouch!)

Anyone out there with any ideas on what to change ?ndd or DSM.SYS

How much can I push through

TIA

Dominic