802.11a/b/g are the basis for MT links, and the consensus seems to be that about 50 customers is the limit for an AP. Since these are collision/detect based protocols, what is the actual limiting factor (collisions, CPU, ). Using a collision base protocol in an outdoor distribution environment in general is not a good thing as most or all customers can not see each other, you end up with the old hidden node problem and a rapidly decreasing efficiency and performance as you ramp up the customers.
Does the limit come from the limited CPU power of the AP’s or is it from the limits of a collision base protocol in such an environment? If it is from the collisions does it make sense (and does it work) to have several AP in the area that are all on the same frequency? and/or to have customers using omni antennas when ever possible? This would in theory enable a high number of units to see each other and thus reduce some of the hidden node issues and thus improve the bandwidth efficiency of the system. Any feed back on this?
What is the impact of having some customers hooking at a 6mb speed and other at 54mb speed? Does this help by distributing the load ?
Then there is NStreem polling, on most of my tests with RB532 and RB112 boards I saw a reduced through put and extremely high CPU usage with NStreem, and as a result have not implemented it so far. How well does this work? And is it more efficient use of my bandwidth due to the polling in a distribution environment. Can I get more customers on the system? Do I need to go to a P4 level ap to make it work?
We are debating on rolling out a few MT distribution sites which have the potential to have 200+ customers with minimum peak bandwidth of 2mb for residential customers. But we have only really worked with P2P links so far.