Hi,
I'm testing 2 RB1100 connected via a 400Mbit/s microwave link (carrier grade) with gigabit ports.

When I start a speed test using UDP protocol from one RB1100 to the other one:

• - if I do not set a speed limit, the highest speed reached is quite wrong (about 60Mbit/s), looks like the test is not capable of reaching the highest speed possible

• - if I set a speed limit (1000M in the picture) the speed test returns on one RB1100 the expected speed the microwave link is capable of (337Mbit/s), but in the opposite RB1100 the speed shown is the limit I set in the speed test!!

I tried setting speed limits of 800M, 400M, 200M and in the opposite RB1100 I used to get the speed set as limit of the test.

This happens with both v4.10 and v4.11. Both RB1100 have the most recent bootloader (2.27).

Bye,
mattia

In case of UDP BW server tries to send packets with 1gbps speed, that is why you see that traffic on outgoing interface even if received is only 300Mbps. UDP do not have windowing mechanism like TCP has to limit transmit speed.

Thanks for your reply MRZ, but I hope you agree that a speed test that returns unreal values is quite useless!

And if I do not set any speed limits, the speed test returns wrong values, so you should agree it's not working properly!

Nothing is wrong, everything is working perfect. You didn't have to set that 1000Mbit speed. The router is trying to meet your set requirement, but your link just doesn't work so fast.

Hi Normis, maybe you didn't noticed I'm using a 400Mbit microwave link: it's a 18GHz carrier grade device and those 400Mbit/s are real and effective, it uses 256 QAM modulation and was designed for our specific purpose, it's not a best-effort link!

So, that's why I say the speed test is malfucntioning: it returns poor results without setting a limit, it's wrong when limit is set.

And I don't know any other speed test tool that returns the limit you set as result of the test... this would be "who wants to be a Millionaire"

Thanks,
mattia

OK, lets go back to beginning.

1) you didn't notice is that there are always additional management BT connection between both ends of the test.

So you send 1000M UDP traffic out and then other end using TCP connection send exact information back how much packets actually made it, if management connection is unable to get packets trough it disconnects test eventually.

2) if you do not apply a limit to BT it is looking at the management connection to determine what is the proper speed, so if your packets start dropping early or management connection have hard time finding way back to source, it can stay on lower values than when statically assigned.

3) you specify ŗemote TX speed" in picture so remote end must send out 1000Mbps - picture shows that it do so, so there are no problems, you need to look how much traffic you are receiving on this router.

4) setting limit 1000Mbps to device that might be possible to handle only 300Mbps is same as DoS attack, so i suggest to set values pretty close to reality.

5) Is this 400Mbs actual throughput or air speed? is it one directional?

6) and LAST: all here know that price is not guarantee that something will work better, so it looks like your mighty Wireless equipment is unable to handle some UDP DoS attack and act poorly under such conditions - i suggest to swap it for MikroTik

Ok, so please delete this topic!

Testing bandwith directly with a utility on the router himself is not good practice because it will load the router and it will not reflect real routing or bridging performance.

Such "on router" testing should be considered like an helper utility for first diagnose, but not like a certification test.


Use a hardware RFC2544 test system connected at each end if you need trusted results.