Link aggregation on device with only 1 physical interface facing the internal switch

[Livy]

QUESTION 1

I have a Newifi D2, which has only 1 physical interface eth0. This device switch configuration is as follow:

• The integrated switch has 6 ports: 0, 1, 2, 3, 4, and 6
• Port 0, 1, 2, 3 are for LAN (vlan 1) and port 4 is for WAN (vlan 2).
• Port 6 is connected to the CPU physical interface eth0
• The CPU does routing between LAN and WAN by using eth0.1 and eth0.2 sub-interfaces. In other words, all sub-interfaces connected to the CPU share a single link speed of 1gbps.

Does it make any sense if I do link aggregation by using 2 physical switch ports? Theoretically the CPU only has 1gbps link to the switch, so bonding does not help right? Except maybe for redundancy?

QUESTION 2

I also have a TP-Link TL-WR740N, which has 2 physical interfaces eth0 and eth1. This device switch configuration is as follow:

• The integrated switch has 5 ports: 1, 2, 3, 4, and 0
• Port 1, 2, 3, 4 are for LAN (vlan 1), with a link sped of 100mbps.
• Port 0 is connected to the CPU physical interface eth0, with a link speed of 1gbps. All VLAN sub-interfaces share this link speed.
• The WAN port does not belong to the switch, but is rather an independent port which leads to another CPU's physical interface namely eth1. This port is also 100mbps, just like all LAN ports.
• The CPU does routing between LAN and WAN by using eth0.1 sub-interface and eth1 physical interface.

In this case, is it possible to do link aggregation using the physical interface eth1 and a sub-interface, let's say, eth0.15? Because they have different link speed: 100mbps vs 1gbps. If it is possible, does the weak CPU likely become a bottleneck?

[dlakelan]

It doesn't make sense to link aggregate anything other than physical interfaces. If you want redundancy use two links to the switch and run spanning tree. One will be up and if it dies the other one can come up

[anon69880279]

for any aggregation of two network links you need two different devices to do this