A Case for Adapting Channel Width in Wireless Networks Ranveer Chandra, Ratul Mahajan, Thomas Moscibroda, Ramya Raghavendra, Paramvir Bahl SIGCOMM Soyoung Park
Contents Background Methodology Main part 1.Impact of channel width 2.Benefits of adapting width 3.SampleWidth algorithm Conclusion
Background A (MHz)2A (MHz) Time/symbol1/A =T S (μs)1/2A = T S /2 (μs) Power/secPP Power/HzPhPh P h /2 Waveform Power Spectrum Density (PSD) A 2A /18 Effect of channel width change
Methodology Only OFDM rates (6~54 Mbps) used in g mode –Atheros-based NIC Channel width is varied by frequency of reference clock: 5, 10, 20, 40MHz –Reference clock affects not only channel width but also timing parameter (SIFS, slot duration …) –Modulation is not affected [ A few timing parameters across channel ]
Experiment environments Controlled emulator setup –CMU’s wireless channel emulator –Two laptops connected through an FPGA FPGA models signal propagation effects
IMPACT OF CHANNEL WIDTH STEP 1
Peak throughput Peak throughput increases as the channel width is increased Shannon capacity
Narrower channel widths are able to withstand greater attenuation Higher delay spread resilience in narrower channel widths Improved SNRResilience to delay spread Transmission range (1/3)
Transmission range (2/3) Improved SNR – In FCC regulation, the total transmission power is fixed – Power per unit Hz decreases as the channel width increases Assuming total power per sec equals P – 10MHz : power per unit Hz = P/10MHz – 20MHz : power per unit Hz = P/20MHz But, noise per unit Hz is same noise transmission power same area = same power [Narrow channel width][Wide channel width]
Resilience to delay spread –The narrower Channel width is, the longer symbol duration is Narrower channel is more resilient to delay spread [Transmitter] Inter-symbol Interference TdTd TdTd Multi-path Transmission range (3/3) [Narrow channel width][Wide channel width] [Receiver]
Wider channels consume more power –Power consumption depends on the clock speed Energy consumption
BENEFITS OF ADAPTING WIDTH STEP 2
In case of Fixed channel width –The higher transmission power, the larger range Tradeoff relationship between range and power conserving In case of Adaptive channel width –The range is increased with less power consumption by narrowing channel width Reduce power consumption & Increase range
The channel width offering the best throughput depends on the distance between the nodes –Hence, we need to adapt channel widths based on the current situation Improving flow throughput
THE SAMPLEWIDTH ALGORITHM STEP 3
Algorithm sketch Major challenge is the size of search space –(# of options) = (# of modulation) Ⅹ (# of channel width) To reduce a search space, two dimension is decoupled –Adapt channel widths periodically –At given width, find best rate using autorate algorithm Because switching channel width incurs more overhead Channel-Width adaptation algorithm – R < α(9 Mbps) : switch to the adjacent narrower channel width – R > β(18 Mbps) : switch to the adjacent wider channel width – α < R < β: switch to the channel width resulting highest throughput in history
For all attenuations, the throughput achieved by SampleWidth closely tracks the throughput yielded by the optimal width Performance evaluation
Conclusion Adaptive channel width –New metric for improving network performance –Optimal channel width is changed based on communication environments and requirements Narrower width: wide range, low energy consumption Wider width: high throughput