[Olsr-dev] Node hand-off and attach speed

Aaron Kaplan (spam-protected)
Wed Mar 26 18:54:16 CET 2008


>
>
> In a situation in which the moving node currenly has an open TCP  
> connection with a remote host, will that connection be dropped, or  
> will it persist?
olsr simple calculates routes, lots of host routes. So the TCP  
connection will be still up as long as you don't switch NAT gateway  
on the way
(but there is a solution for that also)

> e.g. if the moving node has someone telneted into it, would the  
> user's connection be dropped, or would the connection be re-routed  
> without issue.
re-routed

> (i'm sorry if this is fairly basic, I'm still trying to determine  
> if OLSR meets my needs).
>
> > 2) Do we know the average connection or transfer time when you  
> have nodes
> > moving from one to another?  To give context, lets say you have a  
> number of
> > nodes that are in fixed locations with mobile nodes (think  
> vehicles) moving.
> > Based on node speed and range, the mobile nodes could move too  
> fast to
> > connect.
> Connection speed depends on the timeout settings of the routing agent.
> A default OLSR router transmits a single HELLO packet every two
> seconds. You need a bidirectional connection, so it takes 2-4 seconds
> to establish a link (more if packets get lost).
>
> If you set olsrd to "0.2 seconds" intervals for Hellos you could
> establish a link within 0.2-0.4 seconds. Of course you pay with the
> higher overhead, but it might be a good idea. Of course you have to
> decrease the TC interval too (default is 5 seconds).
>
> > If anyone has any insight into this I would really appreciate  
> it.  I'm at
> > the preliminary stages of development and testing and am unable  
> to test
> > these 2 issues.  They are effectivly showstoppers (along with  
> securing the
> > network).
> Maybe you should tell us a little bit more about your scenario...
> what's the maximum speed of the nodes relative to any other node ?
>
> The very worst a relative speed of about 225 kmph, but that would  
> be exceedingly rare.  Most likely the maximum would be aproximately  
> 112kmph and more normally it would be 50kmph and slower (i.e. motor  
> vehicles with both stationary and mobile nodes).  Obviously range  
> of the wireless transmitters comes into play when we start talking  
> about the very fast nodes.
> In addition, I would like to lock down the network with some kind  
> of encryption for the data and with a shared private key for adding  
> new nodes.  I have read some white papers hinting that WPA cannot  
> be used in OLSR networks.
>
why not? basically anything that supports encryption in ad-hoc mode.

> Henning
>
> I appreciate your time.  If these issues can be mitigated, I think  
> OLSR would be a perfect fit and I hope I can contribute  
> meaningfully to the deamon.
>
do so :) the best is testing testing testing... we do have some  
experimental releases

a.






More information about the Olsr-dev mailing list