| Thread Links | Date Links | ||||
|---|---|---|---|---|---|
| Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
Note that there is a distinction between detecting that a new neighbor of a node has appeared and that a neighbor has disappeared. When a neighbor disappears (such as due to a span break), this should trigger a protection event but does not need to trigger full topology discovery. Each node already has a knowledge of the complete topology and need only make a modification to a single span in that stored topology, in this case. (This enables steered protection to work in well under 50 ms in metro area cases.) When a new neighbor appears, however, that must trigger new topology discovery. So the triggers for the protection and for the topology discovery are distinct.
-- Jason
-----Original Message-----
From: Denton Gentry [mailto:denny@xxxxxxxxxxxxxxxxxx]
Sent: Friday, May 04, 2001 9:07 AM
To: stds-802-17@xxxxxxxx
Subject: Re: [RPRWG] Topology Discovery Timer.
> I think there are two implementation to send Topology Discovery
>packet on the ring. One implemenation is "periodical" and another
>"need basis" (Ex: wrap detection)
>
> Should we implement both? How do you think about this?
To get to 50 msec protection, topology discovery will have to be
triggered whenever a significant event occurs.
A topology mechanism which only runs when needed sounds great in
theory, but in practice is likely to be quite fragile. We should
periodically run the topology discovery.