I have run several throughput tests and i have some questions:
Why did you build such great board with 10 Gbit when maxim throughput it’s only 2.4Gbps?
If i connect 2 hosts into mikrotik rb1200 i obtain 1GB throughput both ways.
I have tested throughput between hosts with 1 cable connected to each other they have 2Gbps throughput both ways.
So i have like 500 Mbps max throughput between 2 ports with mikrotik and even more i it’s not a symmetric upload/download bandwidth but more like 850Mbps download / 15Mbps upload
Going farther after configuring VPLS i have like 800-850Mbps throughput one way so it’s like i will always have ~~400Mbps both way on 2 ports connected between 2 mikrotik routers.
i have PC-----Mikrotik-vpls-----------------vpls-Mikrotik----PC
At ~~450Mbps both way traffic cpu load it’s almost all the time 95-100%
Conclusion i have only 450Mbps max troughput through VPLS tunnel on a switch with 2.4 capable throughput with 10 x Gbit port
You shoud definitely add new hardware capable of large throughput with 2Gbps trafic both way /port
I install mikrotik OS 5.5 on a intel board server with xeon x3430 with 4GB DDR3 RAM.
We have test several ethernet chipsets from your supported hardware list and there are no capability of increasing MTU L2/L3 over 1500bytes even if with linux we could increase it to 7000 bytes.
It is very important this thing because if we cannot increase MTU on other platforms other than router board, we have no choice and no hope to have bigger throughput than 400-450 Mbps on a vpls tunnel with any mikortik OS.
6. My question is simple: How do i get bigger throughput with mikrotik ROS?
What hardware did you test and what was your results about throughput? (any kind of hardware not only router board)
Forwarding is done by software so it is CPU and bus limitation. If ports are configured as switch ports, then you can have wire speed on all ports, because forwarding is done by switch chip.
This is not really true. You can increase MTU as per max driver supported size also on x86. You can’t see the L2MTU of the specific interface but it doesn’t mean that you can’t have an mtu bigger then 1500.
In my experience the big deal is to mix RB and X86 in an mpsl-vpls enviroment. In this case you have to fine tune the MTU and (rb side, the L2MTU).
I think to help MT should implement, on OSPF, the ip ospf mtu-ignore .
MTU ignore for OSPF is dangerous as hell. There’s a reason the protocol requires that the interfaces between two neighbors have matching MTUs - if an LSA, LSU, or DBD sent are larger than the MTU on the other side, you’re going to have problems. This is particularly nasty when it all works fine in your lab because you have a small lab network. Then you put it into production, and it still works fine. Your network grows, and eventually, two years later, you add another link to the OSPF area, the MTU is exceeded, and you’re troubleshooting for hours why something that was stable suddenly failed.
If you have an MTU mismatch problem preventing OSPF neighbors from establishing an adjacency the better option is to match the MTUs on both sides.
I don’t need ospf. We have a L2 network and we add several mikrotik for MPLS purpose. But we neet greater throughput as we serve 500/1000Mbps per VPLS tunnel. We cannot obtain this speed with router board. Only with X86 platforms.
We tried to change L2MTU but an error appear that we can’t change it. Also it didn’t work from console