Very bad Speedtest.net results through Mikrotik equipment

Hi.

In my home network I’m using Mikrotik CRS125-24G-1S as a router gateway for my UPC cable link.
The CRS-125 is connected to the ISP cable modem through a 1Gbps link, and the cable modem has a 500 Mbps uplink to the ISP network.

However, when I perform a speed test using Speedtest.net web site the results are very bad (through the Mikrotik):

  • test performed from computers behind the Mikrotik: Download: 20 Mbps; Upload: ~ 25 Mbps
  • test performed with the same computers connected directly to the cable modem (skipping the Mikrotik): Download: 320 Mbps; Upload: ~ 25 Mbps

While testing through the Mikrotik, no IP firewall filter rules are active.
The problem can be reproduced from multiple computers connected through the Mikrotik.
The problem can be reproduced with other Mikrotik equipments as well (2011UiAS-2HnD gives the same results).

Is there any explanation for this behaviour during the speed test?
Thank you.

Looks like bad configuration or cable speed negotiation problem. Investigate it more… What cpu does?

I’m using the default configuration, nothing fancy. All ports in a switch (eth24 is master, eth2->eth23 are slaves), and eth1 is connected to the ISP cable modem.
Port negociation is done correctly at 1Gbps, and during the test the CPU is idling at 75% average.

interface ethernet monitor ether1-gateway once 
                        ;;; ISP uplink
                      name: ether1-gateway
                    status: link-ok
          auto-negotiation: done
                      rate: 1Gbps
               full-duplex: yes
           tx-flow-control: no
           rx-flow-control: no
               advertising: 10M-half,10M-full,100M-half,100M-full,1000M-half,1000M-full
  link-partner-advertising: 10M-half,10M-full,100M-half,100M-full,1000M-full

interface ethernet monitor ether4 once 
                        ;;; Interface where test PC is connected to
                      name: ether4
                    status: link-ok
          auto-negotiation: done
                      rate: 1Gbps
               full-duplex: yes
           tx-flow-control: no
           rx-flow-control: no
               advertising: 10M-half,10M-full,100M-half,100M-full,1000M-half,
                            1000M-full
  link-partner-advertising: 10M-half,10M-full,100M-half,100M-full,1000M-full

tool profile  
NAME                    CPU        USAGE
ethernet                all         4.5%
console                 all         0.5%
firewall                all         7.5%
networking              all         7.5%
winbox                  all           0%
upnp                    all         1.5%
management              all         1.5%
routing                 all           0%
idle                    all        76.5%
bridging                all           0%
unclassified            all         0.5%

You have some kind of problem.
This is my speed on wireless on my RB951ui-2hnd that replace the 2011 because it had no wireless.

can you post your queue tree?

I have no queues defined. I’m using the default configuration that came with the equipment.

The thing is that I can reproduce the problem with another equipment too (RB2011), also using the standard settings coming with the router (no firewall rules, only NAT active).

What could be the problem in this case?
I can’t blame the cable (between the modem and mikrotik) either, because I’ve used the same cable to connect a PC directly to the cable modem (instead of going through the mikrotik) and the test was good.

Thank you.

Since you have exactly 20.00Mbit down, it seems you do have a queue somewhere. Put your full “/export” output here please

I don’t have 20.00 Mbit result all the time. It fluctuates in the 20-22 Mbit range. It was just a coincidence that on that test the result was exactly 20.00.

Anyway I’ve attached the result from an "/export hide-sensitive ".

Thank you.

stupid question whats the routing speed on the CRS, last i recall it’s a switch with a pretty crappy CPU and not meant for routing in the least, or thats what someone told me previously, it’s made to be used as a switch that happens to have ROS on it…

Should really be using a CCR or RB unit for routing…

But as i see you also used an RB… are you sure both sides of the ethernet are negotiating correctly? are you seeing 1000 full duplex auto negotiated on BOTH sides duplex mismatches can be a bi**h

I don’t know about the routing speed of the CRS, but I like to think that it’s more than 20 Mpbs. :slight_smile:
The CRS and the RB2011 have the same CPU from what I know so them having the same performance during the test make sense somehow.
Do I really hit the routing speed limit at around 20 Mbps with these devices? Anyone?

I didn’t check the speed negotiation on the cable modem, but I doubt that’s the issue (any PC connected directly to cable modem negotiates at 1000Mbps, and the speed test is good - at least close to the 500 Mpbs).
Also I don’t know if the cable modem shows somewhere the link speed. It runs a firmware customized by the ISP. I’ll check though.

I have tested also with two cable modems (one from Cisco and the actual one from Ubee) and the behavior was the same with both. That’s why I think that it’s not the cable modem to blame here.

So no ideas guys?
Did I hit the routing speed limit at ~ 20Mbps with these Mikrotik equipments? :smiley:
If yes, then it’s very sad!

Hello,

try to force the duplex on your gateway interface.

Ok. I will test this when I get home in the evening. However, my hopes are very low. :frowning:

Other ideas?

Hi There,

I had the exact same issue on UPC and my CRS125. problem was solved by changing the queue type from hardware-only to default-ethernet. Works like a charm.

See: http://forum.mikrotik.com/t/lower-isp-speeds-after-swapping-450g-to-crs125-24g/85665/1

Gr Alex

I wanted to sugest that but he has the same problem wint 2011 also. And 2011 don’t have that problem reported…

So guys, here it is:

First time I’ve changed the queue-type for “ether1-gateway” interface, and did a speed-test.
The result was exactly the same (attached change_queue_ether1_gateway.png).

Afterwards, I’ve changed the queue-type also for the master port “ether24-master” and did a speed test again.
The result was surprising. :slight_smile:
(attached change_queue_master_port.png)

So it seems that the default queue types are not the best options. I wonder if this is applicable only for UPC or other ISPs too.

Thank you very much for helping!