Thoughts on Meru and others
In my opionion the Meru single channel architecture does not scale well. Make sure that if you start with a coverage model that you are aware of the costs if you want to move to a higher capacity situation.
Scaling is done “vertically” (by stacking channels) in Meru’s solution, which means that more channels must be deployed across the entire site in order to scale. Since a Meru AP only has two radios, moving from 2 channels to 4 channels means doubling the number of APs, hence doubling the cost of their solution. While vertical scaling will work with the right physical deployment, it’s extremely expensive to achieve in comparison to a multiple channel architecture.
The multiple channel approach is what 97% of the industry is using including market leaders Cisco, Aruba etc etc. Meru claim to be a generation ahead of these guys but the term 4th generation comes from an old Gartner report that stated “Two basic controller architectural choices remain, which Gartner calls third and fourth generation. Both are viable,” but Gartner then goes on to say “The fourth generation is not necessarily advantageous". Note that Meru don't use this last comment in their marketing.
Single channels are positioned as the great saviour of wireless co-channel interference - in reality this is to say that all the other wireless networks out there using multi channel are not working properly - I mean come one!
Someone here mentioned that Ruckus's ability to power down APs to reduce co-channel interference could leave dead spots in the wireless coverage. This is classic Meru marketing talk. If a system is globally aware of all APs it's able to power down intelligently to avoid interference and the advantage is in the ability to power up APs too to fill in any coverage gaps created by AP outages and automatically select channels that do not conflict.
The real future of wireless networks lies in the distributed control plane. Removing the controller from the core network removes a single point of failure (only solved by doubling up controllers), it removes the huge demand on your network infrastrcuture that is created by making the traffic u-turn from the edge through the core and back and allows for faster connections, more resilient networks and a cheaper and more linear models for scalability. Look at the controllerless vendors out there is my tip (and I don't mean Xirrus as they have controllers in every array!).