Intel Nic Teaming

[Noah Hamilton]

Hi,

I have a 2003 windows server that is running Intel Nic Teaming. We are
running a streaming server that will once or twice a day retransmit packets
to a few of my streaming customers. These customers of course at this point
have to reconnect to the stream. I am wondering if the NIC teaming is
causing this issue. I have been unable to detect any other packet loss in my
network besides this server??? Any suggestions?

Thanks.

[Ryan Hanisco]

I have seen this happen before if there is an issue with packet sequence
arriving to the client. Usually with load balancing teaming, you'll have the
NICs plugged into different switches. This isn't a problem if they are just
different blades on a switch chassis, but when they are fully seperate pieces
of equipment they can get out of order and retransmit.

Also remember that while they act as a single NIC at layer 3, they both have
seperate layer 2 addresses. This means that your ARP table will have one IP
address pointing to two MAC addresses -- to resolve this, the switch will
prune one.

So things to look for:
-- ARP timeouts that would refresh the table and make it flip inbound MAC
addresses
-- sequence problems
-- Windowing and fragmentation problems

Also, you didn't mention what kind of stream. If its multicast, that's a
whole different kettle of fish!


--
Ryan Hanisco
MCSE, MCTS: SQL 2005, Project+
Chicago, IL


"Noah Hamilton" wrote:

> Hi,
>
> I have a 2003 windows server that is running Intel Nic Teaming. We are
> running a streaming server that will once or twice a day retransmit packets
> to a few of my streaming customers. These customers of course at this point
> have to reconnect to the stream. I am wondering if the NIC teaming is
> causing this issue. I have been unable to detect any other packet loss in my
> network besides this server??? Any suggestions?
>
> Thanks.
>
>

[Noah Hamilton]

This is an internet radio stream. Thanks for the reply.


On 5/13/07 11:49 AM, in article
AF5B5123-E4D6-434E-9DE0-F806151386D8@microsoft.com, "Ryan Hanisco"
<RyanHanisco@discussions.microsoft.com> wrote:

> I have seen this happen before if there is an issue with packet sequence
> arriving to the client. Usually with load balancing teaming, you'll have the
> NICs plugged into different switches. This isn't a problem if they are just
> different blades on a switch chassis, but when they are fully seperate pieces
> of equipment they can get out of order and retransmit.
>
> Also remember that while they act as a single NIC at layer 3, they both have
> seperate layer 2 addresses. This means that your ARP table will have one IP
> address pointing to two MAC addresses -- to resolve this, the switch will
> prune one.
>
> So things to look for:
> -- ARP timeouts that would refresh the table and make it flip inbound MAC
> addresses
> -- sequence problems
> -- Windowing and fragmentation problems
>
> Also, you didn't mention what kind of stream. If its multicast, that's a
> whole different kettle of fish!
>