I am setting up my first mikrotik routers, and when running iperf through them the speed is good in one direction, but only .5M in the other. If you bypass the router its great. Tried 2 different boards. Using default config and just trying to use them as a simple switch for now to verify performance. Have 5.2 and 5.3 in stalled on 2 different 450G boards doing the same thing. I havent added any firewall rules or anything. Any suggestions? Want to make sure board can support 100M throughput in both directions when using as a switch through bridge interface.

450Gs have a built in switch chip that can bridge traffic at wire speeds without punting it to the CPU at all. That would probably be a better choice than software bridging in CPU. Though it'll still get a heck of a lot more than .5 Mbps in that configuration.
Simply configure the master-port of an ethernet port to be the other ethernet port you want to switch with. Assign an IP to the master port, and both ports will be on that network and talk to each other at wire speed.

Thanks that solved the speed issue, but one of the things we wanted to do was use the split horizon feature available under bridging to isolate all ports and only allow ports 2 - 5 to communicate with port one which is our backhaul and not to each other. Also upgraded to 5.4 before trying your solution and that didn't solve the speed issue. With your solution I removed the ports from the bridge, so no ports were members, and added port 1 as the master port to all other ports. Any other ideas on how to fix the speed and allow port isolation? Thanks

They normally do a lot more of simple bridging traffic.

I would recommend doing a system reset, remove default config.

Disable connection tracking, add bridge, and ports to bridge, enable horizon and test.

You should be able to achieve +500Mb tcp.

Best regards,

I tried that with the same results. Performance through the bridge is still only good in one direction, and .5M in the other, yet works great when using it as a switch. Any other ideas?

I think i missed something,

Are you doing a dual mode tcp iperf test?

No I test in each direction independantly. I also use other speed tests as well, all with the same result, once I start using the bridge interface, it limits flow in one direction. I did notice there were issues reported in the 5.3 code on the 450 boards, but has anyone tested the 5.4 on a 450G to verify it is actually fixed? Thanks...

Bump

Hi,

I have two questions:
Which direction is problematic? - In my case it was Upload (LAN to WAN)
Are you testing on local network (low latency 1-3 ms) or high latency WAN (Internet)? - In my case high latency causes extra issues to the throughput.

I tried that with the same results. Performance through the bridge is still only good in one direction, and .5M in the other, yet works great when using it as a switch. Any other ideas?

It is the upload going through the bridge. I can reverse cable through different ports and always the same results. As soon as I remove the bridge and use as a switch by configuring master port it works great so this would indicate to me it isn't the hardware issue they were talking about in 5.3 and 5.4. I only have 450G boards to test and they all do the same so I don't know if it is a software issue or a hardware issue. I have tried everything with the same results.

I have the same problem. I turned on the "switch all ports" in the menu switch. We only use the routing. Without switching the communication between other ports and ether1 bad.

Please test with v5.4. If results are much lower than official throughput test results http://routerboard.com/RB450G, send report to support@mikrotik.com. Include supout.rif file (http://wiki.mikrotik.com/wiki/Manual:Su ... utput_File) and detailed description on how this test is done.