Two NICs under WinNT 4.0 system

J

Thread Starter

Jamil, Babar

I am using two network cards with two different IP addresses in my WinNT 4.0, TCP-IP and Ethernet based system. Is there someone out there who has
done this before and would like to share this experience with me. I have a few specific questions, such as:

1) How does one specify one card as the primary and the other one as the backup card?
2) Does "NT" decide about switching over to the other card when one fails or is it done via another software installed on top of NT.
3) Etc...

All the Servers, workstations and PLCs in the system have dual NIC cards.

Thanks in advance:
 
Here's my understanding of the answers to the questions

There is no primary or secondary card.

The "switching" over when a card fails is a high level software function. For instance if you are mounting file shares and the primary card fails,
sometime later after the file share timeout expires (a minute or two), you can try to reconnect the drive and it may use the secondary card. This is a netbeui feature I believe.

If the software is attempting a TCP connection to a PLC, for instance, it will try to attach to the PLC based on the subnet masks. That is the
connect request will go out one card. If the subnet mask is the same, this will be card 0 in the network card list. However, if the cable is pulled out of card 0, TCP will still continue to route packets to it and hence a failover to card 1 will not occur. You could at this point manually change the TCP routing scheme, using the ROUTE command (using the route command
improperly is also a good way to completley foul up tcp)

Two ethernet cards are used, typically, in NT to allow a computer to communicate on more than one distinct subnet. For example, the office lan
and the control lan. In this case the subnet masks control whether TCP should go out the office lan or the control lan based on the IP address of the target computer.

In summary, placing two network cards into a computer does not give you network redundancy, it only gives you the opportunity to have network
redundancy if the higher level software support it.

The way I see ethernet redundancy done is two have two distinct networks. Each computer and device is connected to both networks and typically there is not a bridge between the two networks. For each PLC you provide two IP
addresses, one on the primary network and one on the secondary network. When the primary fails the software switches over to the secondary.
Similarily for the host to host communication (e.g. viewer to server, server to server) the traffic is routed across the available network. This is the approach we took in our software. While winsock provides the ability to use
multiple ethernet cards, the high level software must be designed to take advantage of the capability. When evaluating redundancy solutions its important to take a look at the failure latency (e.g. how long does data stop flowing when I have a failure.) Certainly performance like Microsofts file mounts are not what you need for a control system.

There are two pieces of information on our web site the may help you understand what types of network redundancy are used in control systems.
While this information is certainly related to our product, they outline the common type of redundancy and how to implement them.

- Redundancy brochure - this is the one page "marketing thing" - less useful.
http://www.gefanuc.com/Cimplicity/pdfs/servredn.pdf

- Redundacy manual - this tells you the details of cabling redundancy
http://www.gefanuc.com/support/cimplicity/downloads/GFK1353D.pdf

_________________________________
Pete Sage
CIMPLICITY Development Engineer
GE Fanuc Automation, NA
[email protected]
www.cimplicity.com
 
J

Jeff P. LeBlanc

> I am using two network cards with two different IP addresses in my WinNT 4.0, TCP-IP and Ethernet based system. Is there someone out there who has done this before and would like to share this experience with me. I have a few specific questions, such as:
>
> 1) How does one specify one card as the primary and the other one as the backup card?

You can't, they operate as seperate devices. Their IP numbers can be on the same network or on different networks, they still operate independently and do not care if the other fails. To the network your machine has two addresses.

> 2) Does "NT" decide about switching over to the other card when one fails or is it done via another software installed on top of NT.

Your application will use the card corresponding to the destination your trying to communicate with. The network stack does this transparently.

Jeff
 
W

Walters Curt L Contr AEDC/SVT

Jamil,

I am using GE Fanuc PLC's and GE Cimplicity S/W which support redundant Ethernet. I have two separate networks going to separate NIC's in each
host PC and PLC. The networks have different subnet masks. The GE software has entries in its configuration for two IP's for each device. It works well.

Curt
 
M

Michel A. Levesque, ing.

Hi,

You can theoreticaly install up to four NICs in a PC running NT. When you install the cards and the TCP/IP stack you can select each card and set up seperate IP addresses for each card. This
process is called multihoming. NT will transmit on all cards if the IP's are on the same subnet.
All cards will receive all messages (again if all cards are on the same subnet). Switchovers are not really done...transmissions go to all cards
regardless of the state of the network. Receptions are handled the same way, all valid receptions are passed up the stack.

Now for the answers to the obvious question of what happens when the IP's are on different subnets. The answer depends on the destination IP. Does the message go thru a router? a gateway?
Usually NT will try to send on the destinations' IP subnet. If this fails the message is sent on all cards.

BTW, I hope that each card is on a physically seperated network. A redundant card is kind of useless if the network itself gets knocked out. Think about that kamikazi forklift driver.
If both cards are on the same physical network, you are only protecting against NIC failures...a relatively rare event.

Hope this helps.


Michel A. Levesque eng., mcp
Directeur Bureau Montreal
AIA Inc.
mleve[email protected]
 
J

Jerry M Baucum

Jamil,

> I am using two network cards with two different IP addresses in my
> WinNT 4.0, TCP-IP and Ethernet based system. Is there someone out
> there who has done this before and would like to share this
> experience with me. I have a few specific questions ...

The closest we've come and might work for you is to have two IP "segments" with each device connected to both. You still need a way to
decide which segment to use, which depends on your software. We controlled the connection software and switched segments for each PLC on
communication failure.

The only way we were able to control which NIC was used was to have different segments. I'll be interested in any one else out there who has a better solution.

Jerry

------------------------------------------------------------
Jerry M Baucum
Bullet Software
3213 Peppertree Pl.
Plano, TX 75074
Tel: (888) 33-Bullet (332-8553) or (972) 633-9111
Fax: (972) 633-9112
Email: [email protected]
Web: http://www.bulletsw.com
 
A

Anthony Smoke

Hello,
I attempted to do this for redundacy with two networks. I had two subnets. This setup did not work. Either Windows NT was not working properly or the software was not handling it correctly. In any event, I was able to use an Intel NIC card that had two connections. These two connections could be setup for redundacy and had two separate IP addresses. If one network failed, then the other would transmit. I cannot find the card information.

If you have any questions please feel free to call me at the numbers in my signature

Best Regards,
Anthony A. Smoke
Project Manager
Siemens Power, Transmission, and Distribution, Inc.
Tel: (919)365-2015
Fax: (919)365-1431
email: [email protected]
 
W

William Hullsiek

The problem lies with TCP/IP Addressing, and conventions. An address is assigned to a network interface, not to a Host interface.

The OSI / ISO based protocols attempted to address this problem by going to "host-based" addressing, where a host had one address, but
multiple network interfaces.

I would suggest working with Sisco systems. They have experience in this area as part of their work on the MMS-Forum and UCA. You should
be able to layer a small and fast TP4 layer protocol on top of IP.

If I remember things correctly from my class with them...

TP4 will send PDU (protocol data units) down the wire on multiple network streams, the receiving TP4 will throw away redundant packets.


William F. Hullsiek
Software Engineer
 
M

Michael Dannhardt

I'm not sure if I agree with the above 100% - the part about "Their IP numbers can be on the same network". When a program needs to send something to a destination, does the operating system not choose the NIC based on the network portion of the IP (as defined by the netmask) assigned to it? I was under the impression that you could not have two cards on the same network (as defined by the netmask and the network portion of the IP).

Can someone clarify?

Thanks and regards to all,
Mike

-----------------------------------------------------------------------------
From: Michael Dannhardt, DASTEC Corporation
Address: 457A Carlisle Drive;
Herndon, VA 20170 USA
Tel: 703 709 0515
Fax: 703 709 0985
E-mail: [email protected]
Web: http://www.dastec.com
 
W

Walters Curt L Contr AEDC/SVT

Anthony & Jamil ,

Regarding using two networks with different subnets & pairs of NIC's:

One other step that I had to do was put the IP's and host names in a host file under the etc directory. I can't think of any other 'oh by the way's. Hope this works for you.

Curt
 
A

Anthony de la Rosa

I use two NIC cards on all of my IndustrialSQL servers. In my application I have two separate network segments and the SQL server is accessible by either netwrok segment simultaneously. I am running Lighthammers AlarmSuite on one Intouch node in one segment and another Intouch node on a different segment and both machines log data to the same SQL server box.


anthony
 
Check out the Intel Pro/100+ dual server adapter. This card comes with a software driver which provides redundancy transparent to applications. The technology is called adapter teaming. Fail over times are 1-3s.

Configure port 1 of the card to connect with switch A and port 2 to connect with switch B. The switches should be connected via a multi-port trunk for extra redundancy.
 
N

Naveen Kashyap

WE have provided Ethernet redundancy for few of our projects. We have used the D-link DFE-570TX
adapter for the purpose. It comes with 4 ports on a single card which can be configured either with 4 different IP addresses or with 2 pairs of redundant ports & in any other combination you would require. of course the switching between ports is carried out by a high level network management software that comes with the card.
The product costs something like USD 40 .

You can get additional info at www.dlink.com
 
you may add upto four NICs on a WInNt 4.0 machine.However the real purpose is solved only if they are all different subnets,else you can not have any routing..all you'd have is redundant nics which may be used for failovers...also..i am not sure whether network traffic cud increase if ther are multiple nics...pls clarify
 
you may add upto four NICs on a WInNt 4.0 machine.However the real purpose is solved only if they are all different subnets,else you can not have any routing..all you'd have is redundant nics which may be used for failovers...also..i am not sure whether network traffic cud increase if ther are multiple nics...pls clarify
 
Top