Extreme SPB Fabric Configuartion

Userlevel 5
Hi Community

I am new to the Extreme SPB Fabric (Avaya). I have some questions around the configuration to implement what the following:

Below is a high level diagram of the network we are trying to configure in my lab:

The requirement from the client is to be able to extend some L2, Multicast services and from edge to edge without touching the core or speaking to the ISP.
Fabric fits this requirement like a glove.

The client will have 4 sites.

These sites connect to each other over MPLS Point to Point VLPS services from a 3rd Party ISP.

In the core on each site the client will be deploying VSP switches. The bigger sites will get VSP8xxx and at the smaller sites VSP4xxx series switches. (Each site will have at least two of these Core VSP series switches to High availability)

On the edge of each site the client will be deploying new x440-g2 switches.

In my LAB environment I am simulating the ISP MPLS VPLS service with 3 extreme x460 switch, configured with VPLS full mesh config.
I also have two VSP8404's and two VSP4850's to simulate the core as per the above diagram.

I started my lab by just connecting the Core VSP's directly to each other and configuring fabric connect on all 4 nodes.
I then added the Fabric Attach config on the Edge x440-g2's and the VSP ports connecting to the edge.

This service works great, I provision an Vlan on the x440-g2 and assign an i-sid.
I do the same on the remote site and the l2 services is extended, edge to edge as expected.

The next step in my lab is to introduce the MPLS WAN connection and configure the Fabric Extend.
However do I need to configure "Fabric extend" in this scenario?
The reason for the question is that If I connect the Fabric connect ports to the VPLS VMAN port, the ISIS adjacencies is formed and the SPB service is extended.
No fabric extend config.

Any comments on recommended on config/design is welcome.


1 reply

Userlevel 4
Hi Andre,

Isn't the BVLAN transported explicitly by your VPLS? Seems so, but with the ISP you cannot always guarantee that the customer will have exactly same VLAN-s available for his point-to-point mesh between the sites.

Please forgive me as I'm not enough oriented with ISP transport standards, but with VPLS i.e. L2 VPN and with other mechanisms, couple of issues have to be taken into account. I would think you can just translate the BVLANs to carrier-VLANs, but in some materials I got I see VPLS scenario with MPLS routers at customer edge.

You will have to transport BVLAN IS-IS TLVs as well as 802.1ah-encapsulated tenant data traffic. The way ISP handles your traffic requires you to use tunneling (VXLAN on-board with VOSS) or perhaps just translate the BVLANs (I suppose that's your case). All of this requires you to also take a look at ISP MTU so your MAC-in-MAC traffic and perhaps encapsulated within VXLAN tunnel, has to fit the MTU. All issues that might occur in the meantime require different approaches. Fabric Extend is really flexible and thus selecting good approach is critical.

With that in mind, some ISP methods (MPLS Pseudowire, PBB E-line) require creating logical IS-IS interface to each destination, with specifying ISP-provided VLANs (could be different per each link) for BVLAN translation.
Some methods (MPLS IPVPN, where MPLS routers on each customer edge will be in different subnet) require again logical IS-IS interface but with different VOSS config aand... ONA perhaps; the Open Networking Adapter is a small device that is going to be renamed from what can be seen on marketing materials. It enables VSP 4Ks for VXLAN encapsulation. Moreover, if your ISP MTUs are not enough for you, you need to fragment and reassembly the frames (normally you couldn't and it's not applicable for L2 ISP situation!), ONA gives you that so then it is a must. ONA is just for VSP 4K. It is really nice device with also different use cases ('IoT defender'), one I cry about is just 1 Gbps connectivity for FE in that scenario where you should have ONA because of the MTU or because of having VSP 4Ks only.
With scenarios like MPLS VPLS/ELAN/VLAN, where you would have MPLS routers within same subnet at each edge, there was slightly different config.
If you could explain how will the ISP operate, we could select appropriate scenario and work on the config for this.

If you work with pure L2 core you need configuration with:
conf t
logical-intf isis <1-255> vid primary-vid port name (or else, name is optional but nice to have) (could be with mlt instead of a port if applicable)[/code]Then you can see the logical interface UP as well as adjacencies:
show isis logical-interface
show isis interface
show isis adjacencies[/code]Logical interface should be set for each point-to-point. With hub and spoke it is easy, with mesh I think you need complete set of those.

If you work with L3 core and with ONA perhaps, it's a little bit more hassle but not that much. I can provide you with more input about ONA if you will work with it eventually. I've had couple of weeks break from VOSS/BOSS configuration so gotta recall things on-the-fly.

Details on MTU:
- L2 service providers should give you at least 1544 B,
- L3 service providers should give you at least 1594 B (total with MAC-in-MAC and VXLAN headers) - if that requirement won't be met, you will get a log message about IS-IS adjacency not established;
- If for L3 service you have smaller MTU than 1594 B, you need VSP 4K + ONA
- MTU cannot be auto-discovered over IP so the tunnel MTU will not be automatically set (on VOSS default is 1950 😎.

Hope that helps,