Hi.
I have the following home network setup:
CRS125-24G-1S acting as a switch/router (10.1.1.1)
RB2011UiAS-2HnD acting as a 2.4Ghz access point (10.1.1.2)
RB911G-5HPnD acting as a 5Ghz access point (10.1.1.5)
I did some speed tests from 2Ghz AP to the 5Ghz AP (and vice-versa) using the CRS125 as an intermediary, and even though the links to the CRS125 are negotiated at 1Gbps rate, the test results were quite disappointing for me.
Unidirectional (receive or transmit) test results were averaging around 320 Mbps.
Bi-directional test results were averaging around 205/225 Mbps RX/TX.
Taking into consideration that the CRS125 is sold as a gigabit switch I was expecting a lot more from it and I’m wondering if there is something wrong with my unit or setup, or this is just the real capability of the device.
By the way, there is no particular configuration of the switching part in the CRS125. I’ve just used the QuickSet and set it as a router, while the APs are set in the same manner but as bridges.
I have attached a network diagram and the test results. All equipments are running RouterOS 6.25.
Thank you.
In fact you test the cpu of RB2011 and RB9111G. CRS has nothing to do with that speed test. To test CRS you should connect two pc’s in CRS and make some transfers between them.
You are welcome.
Note: Bandwidth Test uses a lot of resources. If you want to test real throughput of a router, you should run bandwidth test through the tested router not from or to it. To do this you need at least > 3 routers > connected in chain: the Bandwidth Server, > the router being tested > and the Bandwidth Client.
So according to the Note from the RouterOS manual I am testing correctly, the router in the middle, in my case I test the speed of CRS125 (at least according to the official documentation).
I think the official documentation does not match the received wisdom on this forum In a lot of the threads you see here about speed testing not giving expected results, the answer is, the speed test on the routerboard will max out the CPU on the RB, not the link under test.
I have the same switch on my desk at work and I shall endeavour to do an iperf test through it between a couple of PCs at some point this week. If I get the same results as you, I shall be disappointed.
Duplicating Mikrotiks test methodology, you will match the scores.
That the circumstances of the test don’t bear much resemblance to any real world scenario for small environments is a different problem.
I’ve repeated the tests and I can confirm that the CPUs of the RB2011 and the RB911 are maxed out during the test.
So it might be as you said, that these routers are not capable to generate the amount of traffic needed to reach cable speed of 1Gbps.
Use a program like TotuSoft LAN Speed Test client and server running on computers connected to then switch. Make sure the switch is configured to keep the network traffic down in the switch fabric and not bringing it up into the cup and bridges for processing.
I think you will find a CRS properly configured to be an impressive device. I certainly have.
What protocol have you used for your iperf test (TCP or UDP) ?
I also did an iperf test with two machines one Debian and one Windows, both running same version of iperf (2.0.5).
I’ve started the server on the Windows machine and used the Debian box as a client. I’ve got quite bad results:
while using TCP: 541 Mbits/sec (TCP window size: 128 KByte)
while using UDP: 490 Mbits/sec (35% datagrams lost)
Ubuntu at both ends. Lenovo laptop one end, HP DL3x0G7 running ESXi the other. Don’t remember which end was the server, but I ran iperf -r so I tested both ways. Didn’t specify protocol, so it would have been TCP. The first test I did was with the switch totally default, the next two tests I did were with ports 18 onwards mastered to port 17. I tested in each group [1-16 and 17-24] and the results were the same. Wasn’t using jumbo frames.
I suggest testing the two iperf endpoints back to back, if you haven’t done so already [otherwise, what are you comparing?].
Finally, just because an endpoint has a gigabit port, doesn’t mean it can saturate it. My last laptop [Acer 7736z, 2010 vintage] would top out at about 500Mbps and that wasn’t maxing out the CPU. Current laptop [Lenovo Y70] easily does >900Mbps.
In a lot of the threads you see here about speed testing not giving expected results, the answer is, the speed test on the routerboard will max out the CPU on the RB, not the link under test.