Here?s a problematic situation I am encountering which must be quite common except that Microsoft?s implementation of TCP/IP turns it into a ridiculous situation.
I?m running a Windows NT server with three NICs.
I have configured two NICs in thefollowing way:
NIC 1 (on motherboard):
IP address 2.0.0.40
Default Gateway: 2.0.0.197
NIC2:
IP address 172.53.3.40
Default Gateway: 172.53.3.9
Each NIC is obviously on a different subnet. Each NIC is also on a separate LAN materialisedby two switches with no links between them. Each physical network has a separate router in the case of logical network 2.0.0.0 the router has the 2.0.0.9 address. In the case of the 172.53.3.0 logical network the router has a 172.53.3.9 address.
I believe this is a legitimate use of multiple NICs on a server. In effect the server collects data from the 2.0.0.0 network, it processes it and distributes the resulting data to network 172.53.3.0
The 2.0.0.9 route is used to allow remote administration of the 2.0.0.0 network constituents
The 172.53.3.9 router allows remote clients to recuperate data having been processed in the 172.53.3.0 network
Microsoft?s implementation of TCPIP allows only for one default gateway for the machine (see MS knowledge base Q157025) other gateways are considered to act as backup if the first gateway fails. This has proved to be an obstacle to the proper implementation of the server NIC configuration described above.
I am sure a lot of you out there have put in place a network distribution similar to the one described. Can anyone enlighten me on how it was done and what the resulting performance was like?
Happy New Year to you all