CMAC : A N ENERGY EFFICIENT MAC LAYER PROTOCOL USING CONVERGENT PACKET FORWARDING FOR WIRELESS SENSOR NETWORKS SECON 2007 S HA LIU, K AI - WEI FAN, P RASUN SINHA D EPARTMENT OF COMPUTER SCIENCE AND ENGINEERING, O HIO STATE UNIVERSITY # Presentation: Jinhyung Lee Computer Network Lab
C ONTENTS I NTRODUCTION C ONTRIBUTION F EATURES E VALUATION C ONCLUSION D ISCUSSION / 17
I NTRODUCTION MAC LAYER DESIGN GOALS FOR WSN L ONG LIFETIME L OW LATENCY L OW MAINTENANCE OVERHEAD H IGH THROUGHPUT E XISTING SOLUTIONS S YNCHRONIZED MAC SMAC, TMAC, DMAC C ONSUME A LOT OF ENERGY ON PERIODIC SYNCHRONIZATION U NSYNCHRONIZED MAC BMAC, XMAC U SE LONG PREAMBLES / 17
C ONTRIBUTION CMAC U NSYNCHRONIZED DUTY CYCLING N O SYNCHRONIZATION OVERHEAD A GGRESSIVE RTS, A NYCAST Q UICKLY MAKE ROUTING PROGRESS C ONVERGENT PACKET FORWARDING A VOID OVERHEAD OF ANYCAST A CHIEVED GOALS E NERGY EFFICIENCY L OW LATENCY H IGH THROUGHPUT / 17
C ONVERGENT MAC A GGRESSIVE RTS A NYCAST PACKET FORWARDING C ONVERGENT FORWARDING / 17
C ONVERGENT MAC A GGRESSIVE RTS A NYCAST PACKET FORWARDING C ONVERGENT FORWARDING
A GGRESSIVE RTS L ONG PREAMBLE MECHANISM OF BMAC H IGH LATENCY B REAKS UP LONG PREAMBLE INTO MULTIPLE RTS PACKETS RTS BURST S ENDER RECEIVES A CTS, IT SENDS PACKET IMMEDIATELY L ATENCY AT EACH HOP COULD BE REDUCED BY HALF / 17 Sender Receiver Sleep Packet Sleep RTS RXCTS RXSleep Aggressive RTS
A GGRESSIVE RTS A SSESS CHANNEL QUICKLY DURING EACH WAKE UP TIME T O ALLOW NODES TO WORK AT A VERY LOW DUTY CYCLE I F RECEIVER WAKES UP DURING THE GAP BETWEEN TWO RTS S MISS RTS BURST / 17 RTS Channel check
A GGRESSIVE RTS D OUBLE CHANNEL CHECK C HECK THE CHANNEL TWICE TO AVOID MISSING ACTIVITIES F OR EACH CHANNEL CHECK, NODES SAMPLE UP TO 5 TIMES B ETWEEN TWO CHANNEL CHECKS, PUT TO SLEEP MODE I NTERVAL MUST BE SHORTER THAN RTS TRANSMISSION TIME / 17 RTS Channel check RTS Channel check RTS Channel check (a)(b) (c) Executed channel check Canceled channel check
C ONVERGENT MAC A GGRESSIVE RTS A NYCAST PACKET FORWARDING C ONVERGENT FORWARDING
A NYCAST PACKET FORWARDING N ODES OTHER THAN TARGET RECEIVER MAY W AKE UP EARLIER C AN MAKE SOME PROGRESS TOWARD SINK R EDUCE LATENCY A NYCAST TO THE ONE CLOSEST TO DESTINATION F ORWARDING SET N EIGHBOR NODES OF THE SENDER THAT ARE CLOSER TO THE DESTINATION P ARTITION INTO 3 SUB REGIONS / 17
A NYCAST PACKET FORWARDING M ORE THAN ONE NODE MAY CONTEND TO SEND CTS E ACH GAP BETWEEN TWO CONSECUTIVE RTS IS DIVIDED 3 CTS SLOTS FOR (R1, R2, R3) P RIORITIZE THE CTS PACKET TRANSMISSION E ACH CTS SLOT DIVIDED INTO MINI - SLOTS E ACH NODE IN THE SAME REGION RANDOMLY PICKS UP A MINI - SLOT / 17 Canceled RTS CTS RTS Sender CTS slot Canceled CTS mini-slot Node in R 1 Node in R 2 Node in R 3 Canceled CTS
C ONVERGENT MAC A GGRESSIVE RTS A NYCAST PACKET FORWARDING C ONVERGENT FORWARDING
C ONVERGENT FORWARDING A NYCAST HAS HIGHER OVERHEAD THAN UNICAST S UBOPTIMAL ROUTES A NYCAST RTS/CTS S WITCH FROM ANYCAST TO UNICAST IF N ODE IS ABLE TO COMMUNICATE WITH A NODE IN R1 C ANNOT FIND A BETTER NEXT HOP THAN CURRENT ONE N ODES STAY AWAKE FOR A SHORT DURATION AFTER RECEIVING A PACKET S YNCHRONIZED WAKE - UP SCHEDULING T IMEOUT / 17
C ONVERGENT FORWARDING / 17
E XPERIMENTS T ESTBED : KANSEI TESTBED 105 XSM NODES 7 X 15 TOPOLOGY, SEPARATION OF 3 FEET I MPLEMENTATION PARAMETERS / 17 CTS-slot length7.488 ms Number of CTS-slots3 Mini-slot length416 μs Number of mini-slots6 RTS packet size44 bytes Double channel check interval10 ms
E XPERIMENTS M ETRICS T HROUGHPUT L ATENCY N ORMALIZED ENERGY CONSUMPTION S CENARIOS S TATIC EVENT M OVING EVENT C OMPARISON CMAC 1%, BMAC 1% CMAC 100%, BMAC 100% / 17
E XPERIMENTS – STATIC SCENARIO / 17 ThroughputLatencyEnergy Consumption
E XPERIMENTS - MOVING SCENARIO / 17 ThroughputLatencyEnergy Consumption
S IMULATION / 17 ThroughputLatencyEnergy Consumption
C ONCLUSION CMAC AGGRESSIVE RTS, ANYCAST, CONVERGENT PACKET FORWARDING S UPPORTS HIGH THROUGHPUT, LOW LATENCY AND CONSUMES LESS ENERGY THAN EXISTING SOLUTIONS D ISCUSSION N O CONSIDERATION OF NODE MOBILITY A WAKE DURATION AFTER RECEIVING PACKET IS SENSITIVE TO PERFORMANCE F OR LOW DATA RATES, CAN ’ T CONVERGE FROM ANYCAST TO UNICAST T OO SIMILAR WITH XMAC / 17
Thank You # 17 CS
Appendix # 17 CS
H OW LONG SHOULD NODES KEEP AWAKE AFTER RECEIVI NG A PACKET ? L ONGER AWAKE PERIOD → LOWER LATENCY B UT LONGER AWAKE PERIOD MAY NOT BE MORE ENERGY EFFICIENT D EPENDENT ON DATA RATE AND NODE DENSITY lambda: packet arrival rate in a Poisson arrival process
25 P ERFORMANCE OF ANYCAST IF LACK OF CONVERGENCE E XPERIMENT SETTINGS : V ARY TRANSMISSION RANGES TO CREATE DIFFERENT NODE DENSITI ES M ETRIC : L ATENCY NORMALIZED BY DISTANCE ( HOPS IN UNICAST ) R ESULTS : CMAC 1% ACHIEVES LOWER LATENCY THAN BMAC 1%