I have one CRS226-24G-2S+ -IN interconnected with a Ubiquiti ES-48-500W switch. I am linking the two devices via parallel 10G links on the SFP+ ports with RSTP enabled. I am running bandwidth tests, first from between two ports on the CRS device. During this test I am able to achieve over 700 Mbps without stressing the CRS CPU at all. When the bandwidth test is done from between the two switches (across the VLAN trunk) I get CRS CPU utilization of 100%. (The CPU on the Ubiquiti device only rises to about 45%.)
All ports on the CRS are slaved to Ether3. Bridge1 (where RSTP is configured) is linked to that port as well.
I’m wondering, since all ports are slaved (including the SFP+ ports) to Ether3, why is there a difference in CPU utilization when connecting two devices internally, Vs. two devices across the VLAN trunk. Other than the physical interface, the only difference in the configuration of the ports is the VLAN tagging which takes place on the trunk.