[net2-wg] new tag test4

[Rodrigo Fonseca]

Hi,

I created a new tag

net2-beta2-test4

due to some fixes to the TreeRoutingEngineP code.
there was a problem in updating the routing info of a node when that 
node was already the parent.

The current issue right now in the TreeRoutingEngine is that there is no 
replacement policy in the routing table.

For the code freeze, the solution is to leave it the same size as the 
Link Estimator table. The actual fix is to just keep
around the k best neighbors, with k being around 3~5, I think. I think 
adding the replacement policy should be done for
the next release.

Also, you can take a look at
http://dju.cs.berkeley.edu/net2/log.10x10.two-roots-3-sends.graph_logical.png
http://dju.cs.berkeley.edu/net2/log.10x10.two-roots-3-sends.graph_physical.png

for a graphical representation of the trees formed on a 10x10 grid 
topology with roots at 0 and 99 (two nodes at the SW and NE corners).
These were just quickly generated with a perl script that parses the 
output of TOSSIM logging. Perl has bindings for Graphviz which are
extremely easy to use.

For mirage I was able to run a test but saw no output from the root 
node, will have to investigate more tomorrow.

Cheers,
Rodrigo



Philip Levis wrote:

> On Jun 10, 2006, at 12:28 PM, Rodrigo Fonseca wrote:
>
>> I just did: created an interface TreeRoutingInspect.nc.
>> I added to the main branch, can you fix the tags?
>
>
> I have created a new tag:
>
> net2-beta2-test3
>
> Please use this. It has Rodrigo's interface and fixes the queueing bug.
>
> In simulation, I see some dropped packets when I have ten nodes  
> sending at 8Hz or so, all of which are due to queue overflows at  
> forwarding nodes. I suspect that if we add ECN (not now, but later),  
> then we're going to see very dependable operation.
>
> I'm going to tag for the code freeze tonight (we've slipped a day).  
> Does anyone have some fixes for the app that let you set the sampling  
> rate?
>
> Phil