Protection switching in PBB

Well for the PBT network, path redundancy is definitely important. As this is evolving, some of the vendors are suggesting a MPLS like label-switched-path setup where an alternate path is made for every traffic carrying path in either 1:1 or 1+1 (G.8031) fashion. As my previous blog goes, it applies to the management plane implementation of a fully managed PBT networks.

Other way of achieving this, is a control protocol. The challenges for a de-centralized control plane is that it has to understand many specific scenarios which can come up in the deployed PBT network. Being the core, the 50ms resiliency is very important. Existing solution of xSTP is way too slow to respond to this particular scenario. Give that the network may be fully or partially meshed, it would become rather complex to converge everytime and having multiple redundant paths, would only lead to a complex topologies even with low density networks. It goes against the conventional STP and STP like protocols. In addition, as soon as a new node is added into the network, reconvergence of the network topology is started because of new paths are now available for the traffic to move, which may be undesirable for the carriers in their core. Yet another issue could be the fault detection and propagation towards the root node or the responsible node which is chosen to take the corrective actions after the fault is discovered. Latency issues and the complexity of the control plane are the drawbacks for this approach.

It is therefore evident that a control plane driven PBT may end up being more messy and complex then a management plane driven PBT network.

However, a management plane driven PBT network has its own problems to solve. Every time a new node is added, it has to be configured through the same management tool which is used for rest of the network. So, a human intervention is more. A failure of the management tool might have its own issue and it always has to be backed with a redundant management plane software, having its own complexity issues.

So for the time being, it seems, G.8031 is the way to go along with Y.1731 for fault detection, unless there is a better alternative proposed in the area of protection and/or loop avoidance for the PBT network within its QoS parameters.