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

[Vigneswaran R]

On 06/13/2013 08:38 PM, David Dahlberg wrote:
> Am Donnerstag, den 13.06.2013, 18:18 +0530 schrieb Vigneswaran R:
>> The idea is to write a plugin to get the second (third etc.,) best
>> routes (for destinations) and keep them in separate routing tables.
>> Later policy routing can be used to route different kinds of traffic
>> through different paths (which are unused otherwise).
> Getting the a "second best" and "third best" routing table from the same
> basic topology is dangerous IMHO. You have to be sure, that
>
> a) everyone on a "second best" path agrees to use the "second best"
>     routing table for a given packet.

I agree. Same policy routing should be applied on all the routers. 
(Otherwise, easily we get into routing loops).

> b) the "x'nd best" routing table is monotonous throughout the
>     network. For Dijkstra and the best route, this is the case
>     (if there are no zero or negative costs involved). For a definition
>     of "second best" I wouldn't be so sure.

We are just thinking of a mesh where we may have 2 or 3 different paths 
between one end to the other. Second best or third best is actually 
alternate routes (could be of same cost or higher cost than the best path).

> If nodes have different views on the network, or your routing table is
> not monotonous, routing loops are likely to occur.

Ok.

> How about not using the "second best route", but calculating a second
> routing graph based on a different metric?

Ok. Though I am not sure how this can be done, we can do some thought 
process on these lines. Thanks.


Regards,
Vignesh