ring topology

[bluesea2010]

Hi,

What is the problem for going a ring topology  instead of full mesh 

Thanks 

[Flavio Miranda]

Hi

 No problem but the topology usually reflect the routing protocol you run.  For example, iBGP requires full mesh and OSPF will consider a ring Topology as Point-to-Point. 

But, if you use ring tolology and use static routing, it willl working with no problem. 

[Jon Marshall]

Because you have indirect connections which mean you are reliant on protocol timers eg. if core 4 fails in your topology above core 1 is not aware of it immediately whereas with a full mesh core 4 would immediately detect the failure. 

Convergence can be a lot quicker if the devices are aware of failures at the physical layer. 

Jon

[Joseph W. Doherty]

What Jon describes is true, as link failure, generally, is detected in 50ms or less.  Also generally, with highly "tuned" higher level keep alive times, I've found it's "difficult" to get less than 250 ms detection time.

However, a full mesh doesn't solve this completely.  There can be issues where the link is up, but logically it's down (usually rare, though).  I.e. you still need to rely on higher level keep alives as a back stop.

[Joseph W. Doherty]

Perhaps the biggest problem is some traffic will need to multi-hop to reach its destination.  For example, traffic between cores 1 and 4 would need to transit either core 2 or 3.

Besides increasing latency (usually no big deal on modern devices within a LAN), such multi-hop traffic can saturate a link between devices.  Consider much traffic between, again, cores 1 and 4, which is hopping through core 2.  Yet core 2 traffic for any other core has to share bandwidth on the core 1 and/or core 4 links.