i want my HAP AC lite 6.41 like simple switch. I have make a very simple config wich works :
nothing else but :
reset configuration with no default configuration
setup a bridge with all ports
dhcp clients on the brigde
it works but i loose the half of my bandwidth only on download (test with ethernet) ?? (90Mb vs 40Mb)
after several test, i assume is due to the performance of HAP AC lite
here’s my test :
configuration for all device nothing else but :
reset configuration with no default configuration
setup a bridge with all ports
dhcp clients on the brigde
all test via ethernet
huawei modem b715s : speedtest : 120Mb down/20 Mb up
just an ethernet cable between huawei and :
Audience 6.41 = great i loose nothing
HAP AC2 6.41 = great i loose nothing
HAP AC LITE 6.41 = i loose between 50 to 70% of my bandwidth down (around 40Mb)
may be due to 100Mb ethernet of hap ac lite (1Gb/s for audience and HAP AC2) ?
Fast ethernet negotiation can fall back to half duplex. Worst case is when one side is half duplex and the other is full duplex.
Mikrotik interface status shows the duplex state of the interface.
If the other side does not answer (eg because full/half negotiation is not implemented, or disturbed negotiation) , then half-duplex is assumed by the negotiating interface!!!
The choice full/half may flap between restarts or over time.
If you see half duplex, you might test without “auto negotiation” on the Mikrotik, and set it fixed to “100 Mbps full”.
If the Huawei does not negotiate it will improve. If the Huawei did negotiate before, it will now select half-duplex on its own interface. Performance will be bad then.
There are other limitations to the max net transfer rate than the 100 Mbps overall transfer rate. What is the packet size ? Smaller packets = more interpacket delays and more overhead.
@dad2312
On your mtik - what speed do you see in ethernet ports status, is it 100 Mbps Full Duplex ? Is Hw. Offload activated on bridge ports (look at Bridge/Ports for big H next to port numbers) ?
Yes it is because of performance of hAP AC Lite, look ate the picture below! With 40-45 Mbps it goes to 95-100% Processor usage! So cannot support more traffic! I have more than 80 of those routers and is the same thing! I use this for clients that need under 20Mbps and need WiFi just for small areas!
I’ve tested in both ways and is the same thing! Btest processes not use more than 5% of processor all other is used by networking! That device stays 1-6% of processor usage just doing nothing!
Just trying to help look around …
If I see your screenshot,ingdaka, I see ROS 6.45.7 but still some traces of the old way of using switch hardware (ether2-master), it’s only the name not the config, but it is probably an upgraded config.
Ether1 is not part of the bridge-port group (it’s not slave), and so usually is at the other side of the firewall. Session tracking and even firewall rules may be triggered, using CPU. Just wondering if this is the case.
With dad2312 I see none of this. (all interfaces are part of the bridge). So the switch chip is doing the work. (Atheros7228). Why is it so slow? Store and forward mode (versus cut through mode)? Switch backpressure because of some half-duplex interface? Loop detection or flow control interfering? If have no idea! What do Rx, TX and status counters reveal ?
My little quick and dirty test setup (ROS versions different and taken off a WLAN-ptp network)
I connected a “hAPac2” - to -" hAP ac Lite" - to - “mAP Lite”. Connection from hAP ac2 to mAP Lite passes over the hAP ac Lite switch.
Only the hAP ac2 has 1Gbps interfaces.
My simple test shows me full speed in all directions.
ouaou bpwl thank you very much for your test, i know definitively now i have trouble with something (i don t know where yet but your test is à real proof)
thanks again !
i test huawei b715s with other stuff (not mikrotik) on ethernet
when other device (like connect tv) have 100 Mb ethernet i loose more half on bandwidh
when other device have 1Gbs ethernet i loose nothing