On 4 Feb 2015 01:50, "Bob Copeland" wrote: > It would also be interesting to know > how much worse "combined" is compared to two gates > with twice the combined bw limit, to see what the overhead in mesh looks > like. Hi Bob,
If I get this correctly, the readings I sent earlier regarding all three scenarios (Gate1, Gate2, and Gate1+Gate2) show gains for when combined link speed is kept equal to the link speed of a single gate (Example, Combined time corresponding to two links of 200 kbps each is still less than a single link of 400 kbps). The table I sent earlier shows that combined is almost always better than a sibgle link of double the bandwidth. > Just to restate what Chun-Yeow said, you probably know this already, but > one thing to watch out for in an actual deployment is if both gates are > in the same coverage area and same channel, then adding more gates won't > buy additional bandwidth since only one of the two gates gets to > transmit at a time. This can potentially be mitigated by operating on > multiple channels, reducing tx power, and spatially separating the > nodes. Or maybe some expensive/crazy beamforming techniques :) We didn't consider the idea of simultaneously using multiple gates. Can a mesh interface jump between different channels ? Or do you mean that we can run subsets of the network on different channels? (Say, 10 nearby nodes on Channel 1 while the other 10 on Channel 3. Will a node on Channel 1 be able to efficiently communicate on both Channel 1 and Channel 3 ?) > Another question is whether you've considered or approached the problem of > arbitrary network topologies being bridged through multiple gates. > Normally we have to run STP on the gates to prevent loop formation in > this case, but STP doesn't work all that well with mesh. We haven't considered crazy topologies yet. Just simpler ones like the experiment described earlier. The bridge interfaces are set up using brctl on the gates. STP is turned on by default. We will try to deliberately set up confusing networks and observe how the network reacts. Regards, Ashish Gupta
_______________________________________________ Devel mailing list [email protected] http://lists.open80211s.org/cgi-bin/mailman/listinfo/devel
