[Olsr-users] Route convergence time: a function of scale?

[Hannes Gredler]

On Tue, Dec 09, 2008 at 06:31:54PM -0300, Breno Jacinto wrote:
| Hello Hannes,
| 
| > |
| > |      My question is how long would be this delay, and if there is any
| > | study or practical experiments that demonstrating how long it takes
| > | for everyone become aware of a joining node. In small scale networks,
| > | of course this isnt much of an issue.
| >
| > thats mostly a matter of the average TC origination in your network.
| > at funkfeuer.at (app. 400 nodes) i get 90% of all the routes after 30sec.
| 
| 
|     Thanks for the information. That gives me an idea of how long
| would it take to converge the routes.
| 
|     But, then it may mean that 10% of that those routes are not
| up-to-date. See, I was wondering if a simple "ping <some ip>" would be
| a reliable mechanism to assure if an IP is available or not. But that
| would be true if, and only if, I could be certatain that all routes
| are converged after a period of time, to every node (like the 30s you
| mentioned).
| 
|     In any case, I'm pretty sure now that flooding an address request
| (and another protocol for that...) is necessary, such as the case of
| Pro-Active Autoconfiguration (PAA -
| http://www.olsr.org/docs/report_html/node179.html), to assure that the
| address is not really being used.
| 
|    By the way, in your funkfeur.net mesh, how are nodes joining the
| network? The configuration remains fully static?

yes, static IP allocation.