Hello, I have a RB5009 that does NAT, PPPoE, routing between some vlans and the CPU load is always <5%.
Today I was doing some iperf tests between two clients: they’re both connected to a switch, the first client is connected via a 2.5Gb ethernet, the second one is connected to a SFP+ port at 10Gb and the RB5009 is also connected to the switch using a SFP+ port.
I noticed that while using iperf from client 2 to client 1 (on different vlans, so the RB5009 is routing them), the CPU use gets higher (about 30%) as I expect due to the routing process.
The speeds that I get in this way are about 1.70Gbit/s to 2.00Gbit/s.
If I try to do the same test from the same vlan for both clients, the CPU usage is back to <5% as expected due to no routing happening.
The speeds that I get in this way are about 2.30Gbit/s to 2.35Gbit/s.
I get those speeds also when enabling fast-track.
My question is, how is it possible that I get that lower speeds while routing if the RB5009 doesn’t event max out on CPU load?
That… did the trick, now I’m getting 2.00Gbit/s to 2.35Gbit/s while routing!!
No cores seems to max out, maybe hitting 80% but not more in this test.
Also, is 2.35Gbit/s the max speed that I can achieve on a CAT6 5m cable?
(I put on connectors myself so I want to be sure I did that correctly and prevent any sort of signal loss)
And iperf3 uses TCP by default, which means a 1500-byte IP packet produces 1460 bytes payload. In theory, if you max out the 2.5Gbps port, you’ll get at most