Chorus: Collision Resolution for Efficient Wireless Broadcast Xinyu Zhang, Kang G. Shin University of Michigan 1 Outline Chorus (broadcast) MAC layer PHY layer Introduction motivation principles Design Analysis & evaluation PHY PER Summary simulation network 2 Motivation: CSMA/CA limitation Traditional CSMA/CA (Collision Avoidance): Principle: listen before talking --- akin to human world Collision: packets overlap at receiver Limitation: Listen without interpretation Collision avoidance in all cases --- too conservative 3 Rationale(1/3): CSMA/CR principle CSMA/CR (CSMA with collision resolution): A new MAC/PHY paradigm Overcome the limitation of CSMA/CA CSMA/CR Principle: Collision caused by packets carrying the same data can be resolved! A D B 4 Rationale(2/3): CSMA/CR advantage Improving broadcast efficiency Taking advantage of spatial reuse and transmit diversity C C A A S D S D B B E (a) Traditional CSMA/CA based broadcast E (b) Chorus, a CSMA/CR based broadcast protocol 5 Chorus: collision resolution based broadcast Chorus A broadcast protocol with (r ) asymptotic latency Broadcast CSMA/CR Encourage resolvable collisions via intelligent sensing and scheduling MAC Resolve collisions via signal processing PHY 6 Chorus: PHY layer Resolve the collided packet by iterative decoding P1 A A B C D E Y Z D B P1 A' B' C' D' E' Y' Z' S=A' + C S --- the received symbol. A’ --- estimated based on A. C = S – A’ Decode two versions of the packet: from preamble and postamble, respectively 7 Multipacket collision resolution: Head packet A B C D E P1 packet P3 A'' B'' C'' D'' packet P4 A''' B''' Tail packet P2 A' Head and tail packet: iterative collision resolution Other packets: hard decoding 8 CSMA/CR: MAC layer Cognitive sensing and scheduling Basic rules in SEND: If the channel is busy, and the packet in the air is exactly one of the packets in the transmit queue, then start transmitting the pending packet. Otherwise, degenerate to 802.11 9 Chorus: CSMA/CR-based broadcast Extension to 802.11 broadcast mode Anonymous and decentralized S 10 Performance analysis Asymptotic broadcast delay (unit disk graph model): (r ) pkt length Lowerbound: r. L D header length Lh D network radius data rate Upperbound: r. Best known result for CSMA/CA broadcast: ( r log n) Asymptotic throughput: LD Lowerbound: 3( L h ) Upperbound: D 3 11 PHY layer performance analysis Achievable SNR: SNR max {SNR1, SNR2} Achievable PER: PER PER1 PER2 Error propagation effect (based on a Markov chain model): While resolving a given collision, the error propagation probability decays exponentially with the error length. 12 Chorus: Network-level simulation Implement Chorus in ns-2 • Simulated application and MAC layers • Analytical model for PHY-layer packet reception Benchmark protocol: double coverage broadcast (DCB) * W. Lou, J. Wu, “Toward Broadcast Reliability in Mobile Ad Hoc Networks with Double Coverage,” IEEE Trans. on Mobile Computing, vol. 6, no. 2, 2007 • Forwarding set selection: remove redundant transmissions • Each node covered by two forwarders (retransmission improves reliability) 13 PDR and delay in lossy networks : reception probability at transmission range Chorus is more resilient to packet losses. 14 Scalability: Chorus is less affected by network size. 15 Achievable throughput: Chorus can support much higher throughput. 16 Multiple broadcast sessions: 17 Conclusion Chorus (broadcast) transmit diversity MAC CSMA/CR PHY spatial reuse Chorus: achieve optimal broadcast performance via a software radio based MAC/PHY. 18 Thank you! Error propagation effect: a Markov chain model G : Max error length L G F data length offset between collided pkts Pe : BER of clean symbols Pbc : Probability that error propagation stops, i.e., the next bit is correct even when the current bit is erroneous. Can be bounded: 0.5 Pbc 1 20 Steady state error length distribution: 1 0 1 (1 Pbc )G 1 Pe Pbc i 0 Pe (1 Pbc )i1, i 1, 2, , G 21 Impact of packet size: 22 Related Work [1/2] Broadcast for 802.11 based wireless ad hoc networks Most focused on forwarding node selection to prevent broadcast storming * W. Lou, J. Wu, “Toward Broadcast Reliability in Mobile Ad Hoc Networks with Double Coverage,” IEEE Trans. on Mobile Computing, vol. 6, no. 2, 2007 * R. Gandhi, S. Parthasarathy, A. Mishr, Minimizing Broadcast Latency and Redundancy in Ad Hoc Networks, ACM MobiHoc’03 * S.-H. Huang, P.-J. Wan, X. Jia, H. Du, W. Shang, Minimum-Latency Broadcast Scheduling in Wireless Ad Hoc Networks, IEEE INFOCOM’07 23 Related Work [2/2] ZigZag decoding * S. Gollakotam, D. Katabi. ZigZag Decoding: Combating Hidden Terminals in Wireless Networks, in Proc. of ACM SIGCOMM, 2008. PHY/MAC layer technique to combat hidden terminals Similar decoding algorithm. Rely on MAC layer retransmission to obtain multiple collided version of the same packets Interference cancellation * D. Halperin, et. al. Taking the Sting out of Carrier Sense: Interference Cancellation for Wireless LANs, in Proc. of ACM MobiCom, 2008 A MAC/PHY layer technique. Only works when one packet has much higher SNR than the other. 24