Versatile Low Power Media Access for Wireless Sensor NetworksIntroductionMAC design approachesClear Channel AssessmentCCA: single-sample thresholding vs outlier detectionCCA resultsLow Power ListeningRadio powerup sequence of operationsLPL check intervalLPL and neighborhood sizeExperimental results: throughputThroughput vs power consumptionEnergy vs LatencyConclusionsVersatile Low Power Media Access for Wireless Sensor NetworksJoseph Polastre, Jason Hill and David CullerTo appear in Sensys 2004Introduction•What is BMAC?–A configurable MAC protocol for WSNs–Small core•Factors out higher-level functionality–Energy efficient•Goals–Low Power operation–Effective collision avoidance–Simple and predictable–Small code size and RAM usage–Tolerable to changing RF/networking conditions–Scalable to large numbers of nodesMAC design approaches•Traditional (e.g. SMAC)–“Black-box” design•Users pre-configure duty cycle•Applications rely on S-MAC to adjust its operation as things change•Designed (and optimized) for a set of workloads that might not always be applicable•Minimalistic (BMAC)–Small core functionality: media access control–RTS/CTS, ACKs, etc are considered higher layer functionality (services)•Applications can turn them on and off–More flexible and more tunableClear Channel Assessment•MAC must accurately determine if channel is clear–Need to tell what is noise and what is a signal–Ambient noise is prone to environmental changes•BMAC solution: ‘software automatic gain control’–Signal strength samples taken when channel is assumed to be free – When?–Samples go in a FIFO queue (sliding window)–Median added to an EWMA filter–Once noise floor is established (What is a good estimate?), a TX requests starts monitoring RSSI from the radioCCA: single-sample thresholding vs outlier detection•Common approach: take single sample, compare to noise floor–Large number of false negatives•BMAC: search for outliers in RSSI–If a sample has significantly lower energy than the noise floor during the sampling period, then channel is clearCCA results•0=busy, 1=clear•Packet arrives between 22 and 54 ms•Single-sample thresholding produces several false ‘busy’ signalsLow Power Listening•Goal: minimize listen cost•Principles–Node periodically wakes up, turns radio on and checks channel•Wakeup time fixed (time you spend sampling RSSI?)•“Check time” variable–If energy is detected, node powers up in order to receive the packet–Node goes back to sleep•If a packet is received•After a timeout–Preamble length matches channel checking period•No explicit synchronization required–Noise floor estimation used to detect channel activity during LPLRadio powerup sequence of operations•Goals–Minimize time radio is on–Minimize number of times radio gets started–Minimize sampling time (stage e)LPL check interval•Single-hop application doing periodic data sampling•Sampling rate (traffic pattern) defines optimal check interval•Check interval–Too small: energy wasted on idle listening–Too large: energy wasted on transmissions (long preambles)•In general, it’s better to have larger preambles than to check more often!LPL and neighborhood size•More neighbors: more transmissions–More time spent receiving packets–Less time left to go to sleep•To find the best check interval–Set the reporting rate–Estimate neighborhood size–Best result: check interval that gives lowest effective duty cycleExperimental results: throughput•“BMAC is about 4.5 faster than SMAC-unicast”–Not as fast when ACK or RTS/CTS is used–Differences less pronounced as # of nodes increases – (why?)–Another issue: BMAC has CCA, thus it backs off less frequently (and perhaps the backoff timer is faster)–What about hidden terminal without RTS/CTS?0 5 10 15 200 2000400060008000100001200014000160000 5 10 15 2000.10.20.30.40.50.60.70.80.91Throughput of a congested channelNumber of nodesPercentage of Channel CapacityB-MACB-MAC w/ ACKB-MAC w/ RTS-CTSS-MAC unicastS-MAC broadcastChannel CapacityThroughput (bps)Throughput vs power consumption•10 nodes in a neighborhood•Data must arrive within 10 seconds•Average power consumption per node•Low data rates: SMAC is better–Very low duty cycle•Power vs throughput–SMAC: linear–BMAC: sub-linear•Reason: SMAC duty cycle must increase–More active periods, more SYNC periods–Cost of resynchronizing?•BMAC: larger preambles at low throughput, progressively becoming smallerEnergy vs Latency•10-hop network•Source sends a 100-byte packet every 10 seconds•SMAC: change the DC•BMAC: choose optimal check interval•Poor SMAC again performs worse...–Reason: sync packets, probability of multiple schedules---less time to sleep0 2000 4000 6000 8000 10000050100150200250300350400450500550Latency (ms)Energy (mJ)Effect of latency on mean energy consumptionB-MACS-MACAlways OnS-MAC Default ConfigurationB-MAC Default ConfigurationConclusions•BMAC appears to be better than SMAC–Easier to tune–Has better channel assessment –Doesn’t use explicit sync packets–Doesn’t use RTS/CTS/ACK if it doesn’t have to–Is smaller and less complex•Are large preambles always good?–Not sure, but:•They do mention ‘cyclic packet transmissions’•They can be turned off: –Enable LPL–Send first packet, disable LPL–Send remaining packets in a big burst–Re-enable
View Full Document
Unlocking...