Partially Overlapped Channels Not Considered Harmful Arunesh Mishra, Vivek Shrivastava, Suman Banerjee, William Arbaugh (ACM SIGMetrics 2006) Slides adapted from Ashwin Wagadarikar, Duke Spectral Bands and Channels • Wireless communication uses emag signals over a range of frequencies • FCC has split the spectrum into spectral bands • Each spectral band is split into channels Example of a channel Typical usage of spectral band • Transmitter-receiver pairs use independent channels that don’t overlap to avoid interference. Channel A Channel B Channel C Channel D Fixed Block of Radio Frequency Spectrum Ideal usage of channel bandwidth • Should use entire range of freqs spanning a channel • Usage drops down to 0 just outside channel boundary Channel B Channel C Power Channel A Frequency Channel D Realistic usage of channel bandwidth • Realistically, transmitter power output is NOT uniform at all frequencies of the channel. Channel B Channel C Channel D Power Channel A Real Usage Wastage of spectrum • PROBLEM: – Transmitted power of some freqs. < max. permissible limit – Results in lower channel capacity and inefficient usage of the spectrum Consideration of the 802.11b standard • Splits 2.4 GHz band into 11 channels of 22 MHz each – Channels 1, 6 and 11 don’t overlap • Can have 2 types of channel interferences: – Co-channel interference • Address by RTS/CTS handshakes etc. – Adjacent channel interference over partially overlapping channels • Cannot be handled by contention resolution techniques Wireless networks in the past have used only nonoverlapping channels Focus of paper • Paper examines approaches to use partially overlapped channels efficiently to improve spectral utilization Channel A Channel B Channel A’ Empirical proof of benefits of partial overlap Link A Ch 1 Ch 1 Ch 3 Ch 6 Link B Ch 3 Link C Ch 6 Amount of Interference • Can we use channels 1, 3 and 6 without interference ? Empirical proof of benefits of partial overlap Link A Ch 1 Ch 1 Ch 3 Ch 6 Link B Ch 3 Link C Ch 6 Virtually non-overlapping • Typically partially overlapped channels are avoided • With sufficient spatial separation, they can be used Link A Ch 1 Link B Ch X UDP Throughput (Mbps) Empirical proof of benefits of partial overlap 6 5 4 3 0 10 20 30 40 50 Distance between the 2 links (meters) LEGEND Non-overlapping channels, A = 1, B = 6 Partially Overlapped Channels, A = 1, B = 3 Partially Overlapped Channels, A = 1, B = 2 Same channel, A = 1, B = 1 5 2 1 0 Channel Separation • Partially overlapped channels can provide much greater spatial re-use if used carefully! 60 Interference factor • To model effects of partial overlap, define: – Interference Factor or “I-factor” • Transmitter is on channel j • Pj denotes power received on channel j • Pi denotes power received on channel i I-factor(i,j) = Pi Pj Theoretical Estimate for I-Factor Channel B Channel A -30 dB -50 dB -22 Mhz -11 Mhz FcA FcB • Theoretically, I-factor = Area of intersection between two spectrum masks of transmitters on channels A and B Normalized I-factor Estimating I-Factor at a receiver on channel 6 1 0.8 I(theory) 0.6 I(measured) 0.4 0.2 0 0 2 4 6 8 10 Receiver Channel 12 WLAN Case study • WLAN comparison between: – 3 non-overlapping channels, and – 11 partially overlapping channels – over the same spectral band • WLAN consists of access points (APs) and clients – AP communicates with clients in its basic service set on a single channel • GOAL: allocate channels to AP’s to maximize performance by reducing interference Why use partial overlap? Consider a case where you have 300 APs Partial overlap Non-overlap 5 channels, 60 APs each 3 channels, 100 APs each 60 100 100 60 60 60 60 100 Worst case Worst case Interference by all 100 APs on same channel Interference by all 60 APs on same channel + little interference from POV channels Why use partial overlap? Consider a case where you have 300 APs Partial overlap Non-overlap 5 channels, 60 APs each 3 channels, 100 APs each 60 100 100 60 60 60 60 100 Worst case Worst case Interference by all 100 APs on same channel Interference by all 60 APs on same channel + little interference from POV channels Why use partial overlap? Consider a case where you have 300 APs Partial overlap Non-overlap 5 channels, 60 APs each 3 channels, 100 APs each 60 100 100 60 60 60 60 100 Worst case Worst case Interference by all 100 APs on same channel Interference by all 60 APs on same channel + some interference from POV channels Channel assignment w/ non-overlap • Mishra et al. previously proposed “client-driven” approach for channel assignment to APs • Use Randomized Compaction algorithm – Optimization criterion: minimize the maximum interference experienced by each client • 2 distinct advantages over random channel assignment: – Higher throughput over channels – Load balancing of clients among available APs Channel assignment w/ non-overlap • (X,C) = WLAN – X = set of APs and C = set of all clients • How to assign APs to these 3 channels? – MUST LISTEN TO THE CLIENTS! • To evaluate a given channel assignment – Compute interference for each client: cf c ( ( x) 1) – Sum taken over APs on same channel since channels are independent – Create vector of cfc’s (CF) and sort in non-increasing order • Optimal channel assignment minimizes CF Channel assignment w/ partial overlap = + • Each client builds I-factor model using scan operation • POV(x,xch,y,ych) = 1 if nodes x and y on their channels interfere with each other • To evaluate a given channel assignment – Compute interference for each client: cf c ( ( x) 1) – Sum taken over APs that interfere on own channel + all POV channels – Create vector of cfc’s (CF) and sort in non-increasing order • Optimal channel assignment minimizes CF Results for high interference topologies • 28 randomly generated topologies with 200 clients and 50 APs – 14 high interference topologies (average of 8 APs in range for client) – 14 low interference topologies (average of 4 APs in range for client) Results for low interference topologies • Using partially overlapped channels and I-factor, clients can experience less contention at the link level. Higher layers have better throughput Evaluating deployment strategy • • • • square area, clients distributed uniformly at random Clients can move around Must ensure that APs cover full physical space APs must be distributed regularly 1 11 6 Avg. TCP throughput Evaluating deployment strategy in non-overlap case 1.0 0.8 0.6 0.4 0.2 3 channels 0 400 600 800 Number of Clients • 3 APs – operating over independent channels 1 6 11 – arranged in equilateral triangle 1000 Channel separation vs. transmission range • hard to deploy a new AP into one of the non-overlapping channels without getting a lot of interference • With channel separation, can get much lesser interference 1 4 7 11 Avg. TCP throughput Evaluating deployment strategy in POV case 1.0 0.8 4 POV channels 0.6 0.4 0.2 0 400 3 channels 600 800 Number of Clients • 4 APs – Operating over partially overlapped channels 1 4 7 11 – arranged as a square – Covering same spatial area as non-overlap case • 4 APs can be placed closer Get greater spatial re-use 1000 The Overall Methodology Wireless Communication Technology Such as 802.11, 802.16 Estimate I-Factor Theory/Empirical Estimated once per wireless technology I-Factor Model Algorithm for Channel Assignment Channel Assignment with overlapped channels Repeated for each wireless network Conclusion • Efficient use of the spectrum can be made by using partially overlapped channels • Proper use provides: – Higher throughput – Greater spatial re-use