Hello!
I have two computers connected to the two rightmost ethernet connectors of the RB433 (as pictured on http://routerboard.com/RB433), and they show 100MBit/s connection status.
Computer A is sending about 1800 UDP broadcast packets per second, each with a size of almost 1300 bytes, adding up to about 2.3 MB/s. Computer B wants to receive those packets.
When connected via the routerboard, 13.5% of these packets get lost before reaching B. When both computers are connected directly or via a small switch (a netgear $20 100MBit 5port switch), there is no packet loss.
Looking at the “Performance test results” of the product page, switching 2k packets/s shouldn’t be a problem for the RB433, right?
What might be the cause for this issue?
Are those two ports switched in hardware, or could this be a software configuration issue?
Thank you!
kernelpnc
Look first to the network cable
Have you updated the latest RouterOs en have you update the bios?
The routerboards 433 are good boards with a good performance
I have not trouble with this
What power supply you use for this
I have two computers connected to the two rightmost ethernet connectors of the RB433 (as pictured on > http://routerboard.com/RB433> ), and they show 100MBit/s connection status.
This is then a Layer3 (routing) connection.
Computer A is sending about 1800 UDP broadcast packets per second, each with a size of almost 1300 bytes, adding up to about 2.3 MB/s. Computer B wants to receive those packets.
UDP is not an equal for performing tests like this. Please kook here for a equal test method likes you
should test your RouterBoard! RouterBOARD testing procedures
When connected via the routerboard, 13.5% of these packets get lost before reaching B. When both computers are connected directly or via a small switch (a netgear $20 100MBit 5port switch), there is no packet loss.
The Netgear Switch is acting on Layer 2 and not like s the RouterBoard in Layer 3 of the OSI model.
Looking at the “Performance test results” of the product page, switching 2k packets/s shouldn’t be a problem for the RB433, right?
No, but more, that all must matching together for a test, to be representatively.
- AS from plisk also named you need propper and good shielded cables
//Is was all my cables changing to SSFT CAT. 6a related of the good shielding//
But for your test a CAT.5e S/SFT cable would also march very good.
- The MTU size must be the same on all devices in this test.
- To test those things out I really suggest to take iPerf or jPerf or NetIO
this would be a really good choice in my eyes.
What might be the cause for this issue?
Please read above.
But for those tests see the testing procedure shown in the link above.
Are those two ports switched in hardware, or could this be a software configuration issue?
I really think it should be like a brand new RouterBoard where only some things
where setted up like open declared in the testing procedure in the link above.
That is comming verry near to the tests MikroTik made, but the following things
you made are not matching nearly this tests.
- UPD instead of TCP
- not the same test tools
- not the same MTU settings at all devices
- not the same configuration
And above all test are made in lab or in a testing environment and never will be
matching the given real life circumstances as we all have in out networks at hom, at work
or plain in the field.
Kreacher,
thank you for your elaborate reply!
I had installed OpenWRT on the RB433 and after experiencing the packet loss, tried another RB433UL (which I think is quite similar) with RouterOS. On that board, no traffic was passed at all (I believe it was in its default configuration). Since I don’t know RouterOS, I didn’t test any further. At least it confirms that there is no layer2 switching by default on these 2 ports.
When I describe the ports I’m using, you say “This is then a Layer3 (routing) connection”. Does this mean that there is no way of layer2 switching with the 3 ethernet ports provided?
Is there a diagram of how the ethernet ports are connected to the rest of the hardware (as on the performance testing page)?
I have tried many different cables. The packet loss disappears when the devices are connected with the same cable, but using the netgear switch instead of the rb433.
MTU size is above 1300 on all devices. The size of the packets is constant.
I understand my scenario (UDP,…) doesn’t match the testing procedures. Still, I’m surprised the board fails to route packets when the CPU id 95% idle.
Since I’m working in an unsupported configuration (OpenWRT), I guess I should just accept that the board/os doesn’t support this application.
Thanks!
kernelpnc