Hi all.

I have just set a bridge + VPLS tunnel.
Main problem is my client is consuming all bandwidth available of my network. The traffic generated is L2. I have tried unsucessfully to limit using simple queues. Also i have tried mangle + ip firewall forcing my bridge to obey ip firewall rules.
Is there any way to limit my clients traffic using bridge filter or per interface limit.?



Best regards

QoS or firewall rules won't work on VPLS. If you want to throttle the traffic you have to throttle on the ingress interface. You could also try it on the egress interface if you don't have access to the ingress.

QoS or firewall rules won't work on VPLS.

?? We throttle bandwidth on VPLS interfaces so I don't understand this statement.

I believe he means that you can't shape / qos once the traffic has been labelled and is being switched through the network. Filtering / shaping / qos etc should be done as the traffic enters / exits the VPLS tunnel (for example, as you say, on the VPLS interface itself).

I don't think that is what he is saying but if it is are you saying you can't mark/queue based on EXP of the VPLS shim?

Any QoS on MPLS/VPLS, based on MPLS QoS tags, has to be handled in L2 (for example NV2 QoS).
L3 QoS (queues), do NOT work on MPLS traffic in RouterOS.

This is due to how packet flow in handled in RouterOS, see this: http://forum.mikrotik.com/viewtopic.php?f=2&t=72736

Nothing in those flows show QoS using EXP is not possible. Both WMM and NV2 can queue using info carried from EXP. If abcwarbot issue is QoS further in to the network it will likely be on an RF link and hence WMM or NV2 should be able to take care however the original post suggested it was to do with applying bandwidth limits on traffic being bridged between an ethernet interface and a VPLS interface which again is easy to do so I am still at a loss to see what isn't achievable??

Yes, true. NV2 will schedule packets based on the EXP bit. But there's little control that can be done here, and NV2 only handles scheduling, not throttling.

What MT said, which I agree with, is that any throttling has to be done on the ingress to the VPLS interface, because once the packet is encapsulated in VPLS you can't effect it other than what, say, NV2 does with packet scheduling. The switching of MPLS packets does not pass any queues or chains.

Rich

Technically, throttling and scheduling rely on queuing If we look at the original post, the issue is that 1 client is consuming all bandwidth available on the network. This could mean either this customer is suppose to be able to use the whole bandwidth across the network and the aim is to throttle this customer when others want to use the bandwidth or the customer is only suppose to have say 10Mbps but he is able to burst to the full network bandwidth. If case 1, he could set a low priority on all this customers traffic so that NV2 or WMM will throttle more his bandwidth should other higher priority want to use the network. If case 2, a combination of packet marking plus queue tree at both ends of the VPLS tunnel should take care of that.