(disclaimer: I was not at this WBM but at the previous one - but I think some things can be generalized, so I dare to give some answers here. Others who were there might want to add )
On Nov 5, 2009, at 6:20 PM, Gus Wirth wrote:
On 11/04/2009 06:49 PM, Linus Lüssing wrote: [snip]
PS: Just out of curiosity and also because the reports are in progress right now, what kind of results are you expecting?
Things I'm curious about:
How was the mesh configured?
What tools did you use to test the mesh?
How does the mesh react to the addition of a node? How long does it take a new node to integrate itself into the mesh?
Many of the protocols tested at the WBM tests so far can be tuned to different timing parameter. So, in theory you can make them pretty "fast" (but you have more communication overhead). There is AFAIK no agreed upon standard setting of comparable timing and config parameters.
BATMAN per see seems to have a theoretical limitation that Juliusz found out: exponential time for convergence in the presence of loss. This result was obtained thru reasoning and logic, not thru tests.
How does the mesh react to the removal of a node? How long does it take the mesh to react to the node removal?
see above
How does the mesh handle interference? Did you have anything to generate interference for testing?
interference is a layer 2 issue. Unfortunately at the last WBM in Paris, I wish I had a spectrum analyzer to actually compare the testbed with the "interference surrounding" it was embedded within.
What tools were used to visualize the mesh? Any images available?
Results of throughput testing for one hop, two hop, etc.
What was the CPU load on the mesh nodes while relaying traffic? What were the nodes?
I'll probably have more questions later.
In Paris, it was *very* hard to get comparable results at all. However it made sense to discuss and improve the respective implementations however :)
a.