[Olsr-users] getting second best route from dikjstra's algorithm

[Vigneswaran R]

On 06/14/2013 01:40 PM, Teco Boot wrote:
> The trick is to converge to something that works well. L2 feedback may play an important role. Give the 2nd best route a chance and find out its L2 behaviour makes sense. Before the 2nd best route is taken, we must make sure there isn't a micro-loop.

Thank you for your points. I need to read about how to make use of L2 
feedback. As of now, I am not able to think of how it can be implemented.


Regards,
Vignesh

> Teco
>
>
> Op 14 jun. 2013, om 09:12 heeft Henning Rogge het volgende geschreven:
>
>> On 06/14/2013 09:07 AM, Vigneswaran R wrote:
>>>    +---- J --------- K -------+
>>>   /                            \
>>> A --- B --- C --- D --- E ---- F
>>>         \                      /
>>>          +-- M -- N -- O -- P-+
>>>
>>>
>>> You're right. In the above scenario, A thinks that B,C,D,E,F is the
>>> second best path. However, as far as B is concerned, M,N,O,P,F is the
>>> 2nd best path. When we apply policy routing uniformly on all the routers
>>> to use 2nd best path (for a particular traffic), traffic from A to F
>>> will flow through B,M,N,O,P,F instead of B,C,D,E,F (and which remains
>>> unused).
>> It might be even worse. Think about a network like this:
>>
>>   B---C---D
>> / \ / \ / \
>> A   X   X   H
>> \ / \ / \ /
>>   E---F---G
>>
>> A might decide that A-B-C-D-H is the best path and A-E-F-G-H is the second best one.
>>
>> But E decides that E-F-G-H is the best one and E-C-D-H is the second best one, pushing the traffic back onto the best path from A to H.
>>
>> Henning Rogge
>> -- 
>> Diplom-Informatiker Henning Rogge , Fraunhofer-Institut für
>> Kommunikation, Informationsverarbeitung und Ergonomie FKIE
>> Kommunikationssysteme (KOM)
>> Fraunhofer Straße 20, 53343 Wachtberg, Germany
>> Telefon +49 228 9435-961,   Fax +49 228 9435 685
>> mailto:(spam-protected) http://www.fkie.fraunhofer.de
>>
>> -- 
>> Olsr-users mailing list
>> (spam-protected)
>> https://lists.olsr.org/mailman/listinfo/olsr-users
>