Orhan Ergun 5 Comments

IEEE 802.1aq Shortest Path Bridging (SPB) uses IS-IS as an underlying control plane mechanism that allows all the links in the topology to be active.

In sum, it supports layer 2 multipath. SPB is used in the datacenter; however, it can also be used in the local area network. In this article, Figure-1 will be used to explain shortest path bridging operation.

leaf and spine


In Figure-1, both leaf and spine nodes run IS-IS to advertise the topological information to each other.

In SPB, IS-IS is used by the bridges to find the shortest path to each other, and it allows the topology to be calculated.

But unlike routing, large scale bridging uses only IS-IS link state protocol for the topological information, not for the reachability information.

This means that the addresses of MAC are not advertised within IS-IS.

Some vendor implementations can also use IS-IS to advertise MAC address information since they only need an additional TLV for this operation. Scalability of IS-IS for the MAC addresses advertisement is questionable for large scale deployment; thus, both BGP for MAC address distribution and IS-IS for physical topology creation might be a good option.

IS-IS is used on the underlying physical network to create a topology for layer 2. Furthermore, overlay multi-tenant networks still use flood and prone learning mechanism, also known as data plane learning, for MAC address information learning.

There are two flavors of SPB (as depicted in Figure-2): SPBV and SPBM. SPBV stands for Shortest Path Bridging for VLAN; SPBM stands Shortest Path Bridging for MAC.

The problem associated with SPBV is very similar to provider bridging. In other words, all the nodes learn the MAC addresses of the end hosts. In short, the scalability of core network is still a problem for SPBV.



Figure-2: SPB-V and SPB-M (Source: cisco.com)

You might be asking, if I use PVST+ or Rapid PVST+ that uses separate topology for each VLAN, all the paths in the network will be used. Yes this is correct, but there are two caveats.

One is that you need to carefully design which VLAN will be used on which link  since spanning tree will block one of the links; as a result, planning this bring about management complexity compared to using single tree for all VLANS, and it may increase troubleshooting time due to complex configuration.

The other is that since the second link will be standby and if the first link goes down, reconvergence takes time; in addition, application traffic running on active link will be dropped during a convergence event. If multipathing is enabled, the secondary link is active and only the traffic running on the primary link will be redirected to the second link. However, this operation is very fast since there is no convergence event at the protocol layer.

If multi-pathing is implemented at the hardware, microsecond level re-convergence can be achieved; if it is implemented at the software within milliseconds, traffic can continue over the second link. But convergence at the protocol itself can be extremely time consuming, especially in the case of spanning tree.

As I mentioned earlier, another version of SPB is SPBM, which is the shortest path bridging MAC in MAC solution. It is very similar to provider backbone bridging at the data plane (PBB encapsulation is used); nonetheless, the shortest path bridging does not use spanning tree as its control plane.

Instead of spanning tree, IS-IS is used to build the topology in shortest path bridging. Thus, SPB supports multipath bridging. Also, MAC addresses are hidden from the core of the network.

For data center leaf and spine architecture (as shown in Figure-1), spine switches do not keep state for the MAC addresses, and they do not know MAC addresses.

Thus, overall scalability of the fabric can be much higher compared to SPBV. Also, overall SPBM provides much higher scalability for the layer 2 networks compared to SPB’s other version, SPBM. All in all,  both of them are far more superior to spanning tree due to layer 2 multi-path support, simple operation for complex topologies, and scalability.


Leave a Reply

Your email address will not be published.