White Space Networking with Wi-Fi like Connecitivy

This paper introduces the WhiteFi, a Wi-Fi like network architecture designed to operate in UHF white space bands. WhiteFi is the first network prototype demonstrating the feasibility of WiFi like networking in TV bands.

The incumbents in the UHF bands are: 1) TV stations with strong variable spatial usage map and 2) wireless microphones which have more temporal varying characteristics. Therefore any white space network needs to address the spatial and temporal variations in the channel availability in addition to exploiting the existence of contiguous chunks of spectrum.

The WhiteFi introduces the following three key contributions in order to address the distinct characteristics of white space networking:

1. Channel assignment algorithm: The channel assignment in UHF bands is much harder since the network should consider all the clients observations . The WhiteFi gathers the observations from all clients including their available UHF channels and airtime usage in these channels by other white space devices and defines a new metric "multichannel airtime metric" demonstrating by simulations and experimental results to be a good indicator of what channel and width should be used for the communication.
   
2. Access point discovery: The paper introduces a novel time domain analysis technique called Signal Inspection before Fourier Transform (SIFT) to enable faster detection of channel width used by the AP and also measure the medium utilization needed by the channel assignment algorithm. SIFT uses the packet duration and time between packet and ACK transmissions to estimate the waveform width as they are inversely proportional (for constant packet sizes). The authors further introduce two access point discovery algorithms based on SIFT: Linear-SIFT and Jump-SIFT and prove their superiority over traditional search models.
   
3. Handling disconnection: To avoid any interference with incumbents, the clients immediately change their channel when detect their presence and start chirping a special signal in an already known backup channel which contains the information regarding their channel map. The access point periodically checks this backup channel and tunes its radio to decode the lost client packets in case needed. Then the network will be able to re-establish itself in a good operational channel.
   

The authors further use extensive simulations using QualNet package and some implementations to prove the accuracy of SIFT algorithm and the proposed channel metric along the overall performance of the system.

The paper introduces many clever solutions to the problems one will see in the white space networking and do a great job of combining them. I felt the disconnection algorithm to be a bit too sensitive to corner cases and sill a bit immature. The need for the AP to switch its channel to decode the packet when the rest of clients are still running can cause significant degradation in the quality of service the users experience specially when it occurs often or there are possible false alarms.

Very interesting paper and indeed worth to be in the syllabus.