Orhan Ergun 2 Comments

Ring topology is used mostly for economical reason. It is very common topology in the service provider access, and it is not so uncommon in Aggregation and Core ( Backbone ) networks as well.

Long haul links are expensive thus in order to provide last mile connectivity in the Service Provider access domain, nodes might be connected to the closest nodes which have similar functionality.

As an example  layer 2 DSLAMs are connected to layer 1 DSLAMs,instead of Aggregation HUB if the HUB is at different location and creating a Hub and Spoke would be much expensive than revenue to support rural area customers. Thus ring topology is created.

It is a business decision and we find a technical solution for the problem.

What are the characteristics of a ring based topology ?

To provide rapid restoration is hard ! In many case require unnecessary complexity !

You heard many times that ring topology creates a micro loop , didn’t you ?

In ring topology is hard to provide a loop free alternate path at the layer 3 and most of the routing protocols suffer from these.

OSPF and IS-IS as a link state protocols support IP Fast reroute with LFA technology.

Unfortunately LFA can’t provide loop free alternate path for prefix or per link. In order to find a loop free alternate path, you need an encapsulation in order to hide destination prefixes from the intermediate devices. It can be GRE, IPinIP, MPLS and so on. Remote LFA supports ring topologies and can calculate a loop free alternate path by using MPLS encapsulation.

You don’t have to use fast reroute of course. But don’t forget, you will not find most of the time equal costs for the destinations in the ring topologies thus if you rely on control plane convergence, in case of a link failure you may not fullfill your SLAs. It depends on your customer requirements.

If you have ring topology in the access, also in the aggregation and core domain in the service provider networks, then it is pointless to have Fast reroute and having 50msec, 100msec convergence only on the Core but not in the Access and aggregation.

You might think that if core links or node fails, majority of the customers will be effected, but if the access nodes fail, only small amount of customers will be effected but please try to see the big picture.

Imagine that service provider has a ring topology at every access location and if you don’t enable sub100 or 50 sec convergence at those locations at all,  you don’t know which individual access location will fail. Thus you don’t have overall very fast converged network.

But it is also true that, you use different access technologies and equipments for different access services. As an example business customers may not use DSL or PON at the access and the access layer devices for those customers are different and the aggregation devices are different in most of the cases too. ( BNG for the residential multiplay customers, Business PEs for the MPLS layer 2 , layer 3 VPN customers ).

If they are using common consolidated aggregation layer infrastructure and share the links with the residential customers, although you may not need strict SLA for the residential customers, business customer may need.

Thus fast restoration may not be critical at the residential access devices and links but it is important at the aggregation layer and Core, since core will be common for all services.

What about fast restoration at the layer 2 ?

Routing protocols have a micro loop ( OSPF , IS-IS ) or blackhole the packets ( EIGRP ) in case of a link failure since they rely on control plane convergence if you don’t enable fast reroute mechanisms ( Even in a ring you have a micro loop ) since the downstream device don’t converge as fast as upstream device and they rely on wrong entries.

Instead if they would remove their primary path entries when they receive the upstream devices down notification and they started to forward the packet to their downstream immediately ( instead, they need to find a new best path ) then you would have sub 50 msec converge in any case.

In layer 2, G.8032 for example works exactly in this way. There is a R-APS (Ring Automatic Protection Switching) messages which trigger the MAC flush on the devices. Whenever the switches receive a R-APS messages they snoop, remove the stale entries and start to forward the traffic immediately towards their alternate path in a ring topology.

 
0.00 avg. rating (0% score) - 0 votes