cancel
Showing results for 
Search instead for 
Did you mean: 

NIC Teaming

Ayman Altounji
Valued Contributor

NIC Teaming

I have two problems, the first is intermittently I cannot print to certain HP print servers from the Server but others are working fine and when I try to ping the faulty HP print servers it times out, however if I force the NIC Team on to the other NIC everything works fine for a while and then the fault reappears.

Secondly, when running Performance Monitor on the NIC's (individually) only one NIC is handling Network traffic, is this correct or is the Network Team failing. I have noticed the following devices are not started yet when trying to start them NT states they are already started:-

Compaq NIC Team Transport (CPQNTAC.SYS)
Compaq SCSI Device Monitor (CPQSDM.SYS)

Server is built as follows: -

NT 4.0 SP6a (Build 1381)
RAID 1
RAID 5
2 NC3123 Fast Ethernet NIC (SLOT 1 BUS 5)
(SLOT 5 BUS 5)
NICS Configured as Compaq Load Balancing Team (Adaptive Load Balancing) CPQNTAC V1.00.01.0018

Many thanks
Jonathan Goulding (MCSE)
Microsoft NT & Exchange Specialist
20:20 Logistics Ltd
1 REPLY
Ayman Altounji
Valued Contributor

Re: NIC Teaming

Jonathan,

If this is a ALB Team, you most likely need a firmware update for the HP Print Server because of how it is handling arp
See Below......
It is possible that the NC31XX adapter can fail to complete a print job when connecting to a HP Jet Direct Interface that is utilizing firmware 7.0.X (or previous) The print job will submit correctly, but it may be canceled or "disappear" without warning during the printing process. Packet traces show that the HP Jet Direct will respond with TCP Acknowledgements to the secondary adapter in the ALB pair. The secondary adapter is setup to send data, but only the primary will receive. For this reason, the TCP session between the two devices is canceled. The fix was to Contact HP Jet Direct Technical Services and request a firmware update. Technical documentation has been released to HP front line support.

2ndly lets say it is ALB and you have a 2 port NIC Team. 1 NIC in the team will process all the Rx (Receive) Traffic. Both NICs are capable of Tx (Transmit) Traffic. The Tx Traffic is balanced between the NICs by "Destination Address" of the Frame. We use the last 4 bits of the Destination MAC address to determine what port to send the Tx traffic down.
Example lets say you have 4 Clients connected to this server.
Client A - MAC Address 005008BDF3360
Client B - MAC Address 005008BDF3361
Client C - MAC Address 005008BDF3362
Client D - MAC Address 005008BDF3363

If you take the last 4 Bits of the MAC Address for Each Client.
Client A - 0000
Client B - 0001
Client C - 0010
Client D - 0011

Now here is how the 2 Port ALB Team handles Tx Traffic
0000 - Port 1
0001 - Port 2
0010 - Port 1
0011 - Port 2
0100 - Port 1
0101 - Port 2
0110 - Port 1
0111 - Port 2
1000 - Port 1
1001 - Port 2
1010 - Port 1
1011 - Port 2
1100 - Port 1
1101 - Port 2
1110 - Port 1
1111 - Port 2

So by knowing this, you should have all Rx traffic on Port 1, because only 1 Port can process Rx, and Tx traffic destined for Client A and C would go out Port 1 and Tx traffic destined for Client B and D would go out Port 2.
Obvioulsy the bigger pool of clients you have, the better chance of a more even Tx load balancing taking place.
If you only have 1 node you only transmit to from this Server, then all Tx Traffic going out 1 port is accurate and acting as designed.
You can't think of it load balancing by load because that is not what actually happens.

Does this make sense?
Do you have any questions?

mark-