CRS 226 Bridge penalty

Rather than hijack the other post, I opened a new one.

ryanhaver wrote:
bridging which sends all packets to RouterOS (CPU) for processing effectively nullifying the increased throughput benefits of 802.3ad


I am new to Mikrotik, I have a CRS226, and I have placed it into bridge mode. I did this because I didn’t need any layer3, no routing, DHCP, VLANS etc. I assumed bridged mode simple disabled the router features and disabling the routing functionality would be a performance boost, not a performance killer. Have I made an error?

The only functionality I need at this point is switching.

Forget bridge and read this carefully:

http://wiki.mikrotik.com/wiki/Manual:CRS_features#Port_Switching

Also, I’ve read in some other post that 802.3ad is still not implemented in CRS, so you can’t use it and benefit of wirespeed switching… yet

CRS do not support dynamic LACP yet, but they can be configured with static link aggreation (IEEE 802.3ad draft v1).
http://wiki.mikrotik.com/wiki/Manual:CRS_examples#Trunking

The documentation seems to say that bridge mode or router mode applies to how data flows between switch groups, am I correct?

If I don’t need to do any trunking or routing, I don’t really want to use bridge mode, nor router mode. I really just want one switch group. Correct?

If I do want to create a multi-port link, aka trunk group, then that group of ports would be a separate switch group with a different master port than the rest. I would then have to bridge these two master ports together, and as this involves the CRS CPU, three would be a negative performance impact.

How am I doing so far?

I verified my CRS226 has a bridge1, and there is a checkbox it is the root bridge. But since I have port 1 as the master port, and all 23 others, plus both SFP+ slots as slaves of port1, I don’t think I have anything going through the bridge, other than management traffic, so I don’t have any non wirespeed port to port?

Can someone please confirm if I am on the right track here.

As I see it, you need to start thinking that CRS is a switch plus a single port (switch1-cpu) routerboard all in a single box. The way to set this up is different from any other non-CRS device, even if you can still do things “the old way” using the CPU of the routerboard. CRS has a switch chip and if you want wirespeed you have to set things up on it.

Right

You can set up a static trunk in the switch chip directly, even if the ports have the same master-port as other ports:

/interface ethernet switch trunk
add name=trunk1 member-ports=ether6,ether7,ether8

Check the example in the wiki for hints on how to set up the other side of the trunk.

As there is no bridge, theres no CPU involved, so you get max speed the switch chip is able to give you, aka wirespeed.

I think that you should remove the bridge, because bridge configuration may have preference over the switch-chip settings.

Thank you!
Based on your advice, I removed the bridge, and everything still seems to function correctly. To be honest, the impression that I had from quick set was you had to pick from router or bridge mode. I knew I didn’t need any routing functionality, so I assumed I must want bridge mode.

Now I am left wandering, typically why would someone want to use bridging on a CRS?

Well, you may need to do some filtering that can’t be done directly within the switch chip, i.e. filter pppoe discovery packets or apply some MAC based Nat rules… very specific scenarios that few “ordinary” switches probably wouldn’t be able to cover anyway