2

I want to get your idea on bonding two NICS on a server that is connected to a single subnet with only one IP address in a Gigabit LAN enviroment.

What is the pros and cons of having two teamed Gibit/s NIC card against having only one single NIC card ?

The performance are really doubled ? Can i really have a new bounded interface that work at 2 Gibit/s ? What about load balancing ?

aleroot
  • 3,180
  • 6
  • 29
  • 37
  • 1
    do you mean bonding? – mdpc Nov 08 '11 at 22:59
  • Bonding or Teaming. – aleroot Nov 08 '11 at 23:03
  • 1
    My question would be whether or not link aggregation would provide any benefit in the first place. Is the single NIC being saturated? If not, then having more bandwidth available by aggregating the 2 NIC's isn't going to provide any benefit. – joeqwerty Nov 08 '11 at 23:51

2 Answers2

2

If your switch supports compatible link aggregation, you will get a few benefits. Theoretically, double the bandwidth is available, but in practice you get a much more modest improvement.

However, you do get failover in the event a network cable disconnects or a network interface (on either end) fails. You don't get failover if the switch fails. These tend to be fairly rare failure modes though.

Link aggregation is really mostly useful on switch-to-switch links.

David Schwartz
  • 31,449
  • 2
  • 55
  • 84
1

If you really are talking about bridging the interfaces, and not instead setting up some kind of bonding/aggregation, then spanning tree should automatically disable one of the interfaces, on either your system or the switch you are connecting too. Bridging will give you some fault tolerance, but no additional capacity.

What you may really want is Link aggregation, depending how your switch and computer are configured you can get additional throughput, but you will pretty much never get the theoretical total combined capacity of all your interfaces.

Zoredache
  • 130,897
  • 41
  • 276
  • 420