Multipathing in SPB networks

As mentioned before, one of the key elements of the data plane in a Shortest Path Bridging network is MAC-in-MAC (Ethernet-in-Ethernet) encapsulation. What it basically means is that whenever an Ethernet frame enters an SPB domain, it is encapsulated in a new Ethernet frame. Here’s a little illustration of how it works just to jog your memory a little bit:

In most cases, the new (SPB) Ethernet header is similar to the standard Ethernet frame header. However, there are two fundamental differences:

Now, the B-VID field is something we haven’t talked about yet. What is it? What’s its purpose? The short answer is multipathing, or the capability to load-share traffic over several paths between the sender (source BEB node) and the receiver (destination BEB node), given that two or more equal-cost trees exist between both points.

B-VID is an acronym for ‘Backbone VLAN ID’ (sometimes also called Base VLAN ID). Although it shares some similarities with a normal VLAN as we know it from 802.1Q, the properties and uses are a little bit different:

Please note: for the sake of simplicity, we're numbering the backbone VLANs 1, 2, and so on. Since this VLAN ID is also 12-bit (just like a normal VLAN), you are in fact free to use any value from 1 through 4095. However, best practice recommends using values in the range between 4050 and 4065. In our configuration example in the previous chapter, we configured two backbone VLANs and numbered them as 4051 and 4052, which are the default values used by Extreme (and formerly Avaya).

Now, let’s assume a simple network with two backbone VLANs, as shown in the figure 3 below. From the IS-IS perspective, there are two equal-cost paths (which in SPB terminology is referred to as Equal Cost Trees, or ECTs in short) between BEB-1 and BEB-2, both with a cost of 30 (assuming that the cost on each link is 10).

Since we have configured two backbone VLANs, IS-IS will program the first equal-cost path in the forwarding database of backbone VLAN number 1 and the second equal-cost path in the forwarding database of backbone VLAN number 2. This is illustrated in the figures 4 and 5 below.

Now, the only question that remains is how will the traffic be load-shared between these two paths? It is very much up to the implementer, but in Extreme’s case, it is done according to the following (a bit rudimentary) traffic steering:

For virtual networks with an I-SID (service ID) that is an odd number (1, 3, 5, 7 … 10001, and so on) all traffic is forwarded on the backbone VLAN number 1, and for virtual networks with an I-SID that is an even number, all traffic is sent over the backbone VLAN number 2.

Please note: the use of two (or more) backbone VLANs is for load-sharing, not for redundancy! A link or node failure will initiate a new shortest path computation and can lead to new ECTs, but it might as well lead to just one single shortest path tree.

So, this should give you a rudimentary look at multipathing in an SPB network. If you think that this is not very sophisticated, you’re probably right. However, it has proved more than OK for all the deployments so far. Additionally, more advanced capabilities are coming, including IEEE 802-1Qca Path Control and Reservation which will allow more sophisticated traffic steering. But of course, the vendors have to implement this.

In the next part of the ‘Shortest Path Bridging for Beginners’