If it's by hostname, then it should be vaguely round-robin, if your DNS
has multiple address records.  All connections for the cient would use the
IP chosen by that clients' resolver libraries.

You could also try using etherchannel.  If set up correctly, you would
simply have two NICs with the same IP.  If you used src/dest port hashing,
you could get a rough semblance of load balancing across them.  This would
require a switch that supports etherchannel, and an OS or device drivers
that supports it.

Another option would be to upgrade to gigabit if you're 100mbit, or to
10gigabit if you're already gigabit.

Finally, you could just run 2 networks, and the second network could be
for a small number of large clients, and the first would serve everyone
else.

On 06.04.06 at 15:50 david.tyree AT SGMC DOT ORG wrote:

Date: Thu, 6 Apr 2006 15:50:41 -0400
From: "Tyree, David" <david.tyree AT SGMC DOT ORG>
Reply-To: "ADSM: Dist Stor Manager" <ADSM-L AT VM.MARIST DOT EDU>
To: ADSM-L AT VM.MARIST DOT EDU
Subject: DSM.OPT file

           I've got a question about the usage of the
"TCPSERVERADDRESS" line in the dsm.opt file on the clients.

           Currently I have the IP address of the TSM server listed in
the dsm.opt file on the clients. We will be implementing a new
addressing scheme for all of our servers in a couple of months and that
includes the TSM server itself.

           In our case the DNS name of the server happens to match the
name of the TSM server instance. Since our TSM server only has one NIC
and thus only one IP address I went ahead and changed the line to show
the DNS name of the TSM server instead of the IP on a few of the
clients. The DNS name of the TSM server will not be changing, only the
IP address. I wasn't sure at the time if the TSM client would work with
the DNS name instead but I since found in the TSM docs it would. And
found that it actually did work just fine.

           The next issue is that we are thinking about adding another
NIC in the TSM at some point in the future to help split the load on the
network. At that point the TSM server will then have two IP's. Any
clients that I want to have come into the TSM server on the second NIC
would have to have the IP of the second NIC in its dsm.opt file. That
part makes complete sense.

           But what about any clients that I have with the DNS name of
the TSM server in the dsm.opt file instead of the IP address? Which NIC
would they connect to? If I absolutely wanted to ensure that the clients
came into the TSM server on the right NIC I would make sure I had the
right IP listed in that clients options file. But what would happen if I
left the DNS entry in the options file?





David Tyree
Enterprise Backup Administrator
South Georgia Medical Center
229.333.1155

Confidential Notice:  This e-mail message, including any attachments, is
for the sole use of the intended recipient(s) and may contain
confidential and privileged information.  Any unauthorized review, use,
disclosure or distribution is prohibited.  If you are not the intended
recipient, please contact the sender by reply e-mail and destroy all
copies of the original message.