Hi everyone,
Just trying to get some thoughts and knowledge from some of you out there who has experienced and tested this thing.
First off, why I got curious about hardware offloading in RB5009, is because I have a bridge interface named “bridge-customer” and multiple ports are part of it.
5 x VLAN interfaces (ether1.100, ether2.100, ether3.100, ether4.100, ether5.100)
and 1 physical interface (ether6)
ether1 to ether5 goes to multiple switches and traffic are sent from those switches to 5009 as tagged VLAN 100.
device on ether6 doesn’t have VLAN capability so untagged frames coming into ether6, gets part of the same L2 domain as ether1.100 to ether5.100.
I run DHCP on top of bridge-customer.
I just noticed that “Hardware Offload” is activated on ether6, but not on the VLAN interfaces. Why I notice this is because when there’s huge traffic received on ether1.100 to ether5.100, the CPU spikes. 500Mbps can trigger 50% CPU usage. Mind you, I don’t have filter rules, ONLY single NAT rule.
I don’t have a way to generate traffic on device attached to ether6, so it’s a question to me whether traffic from that hardware offload activated port will contribute or not to CPU spike.
Website says that bridging and routing forwarding speed of RB5009 is the same. So, it makes me think, if I do inter-vlan routing in a way that all ports are part of bridge (so ports are HW offload activated), TAG VLAN and enable VLAN filtering, will it make a difference in terms of performance?
