B-MAC Tunable MAC proocol for wireless neworks Summary of paper Versaile Low Power Media Access for Wireless Sensor Neworks Presened by Kyle Heah
Ouline Inroducion o B-MAC Design of B-MAC B-MAC componens Evaluaion of B-MAC Summary
Inroducion o B-MAC B-MAC = Berkley Media Access Conrol A simple carrier sense media access proocol Link-access proocol only Exposes parameers o higher nework layers Tunable media access insead of a black box
B-MAC Design Objecives Principles Reconfigurable MAC proocol Flexible conrol Hooks for sub-primiives Backoff/Timeous Duy Cycle Acknowledgemens Feedback o higher proocols Minimal implemenaion Minimal sae Primary Goals Low Power Operaion Effecive Collision Avoidance Simple/Predicable Operaion Small Code Size Toleran o Changing RF/Neworking Condiions Scalable o Large Number of Nodes
B-MAC Feaures Reconfiguraion and conrol of link layer proocol parameers Acknowledgemens, Backoff/Timeous, Power Managemen, Hidden Terminal Managemen (RTS/CTS) Abiliy o choose radeoffs knobs Fairness, Laency, Energy Consumpion, Reliabiliy Power consumpion esimaion hrough analyical and empirical models Feedback o nework proocols Lifeime esimaion Mechanisms o achieve nework proocols goals
Oher MAC proocols S-MAC Ye, Heidemann, and Esrin, INFOCOM 2002 Synchronized proocol wih periodic lisen periods Black Box design Designed for a general se of workloads User ses radio duy cycle SMAC akes care of he res so you don have o Inegraes higher layer funcionaliy ino link proocol T-MAC van Dam and Langendoen, Sensys 2003 Reduces power consumpion by reurning o sleep if no raffic is deeced a he beginning of a lisen period Node 1 Node 2 sync lisen sleep lisen sleep sync sync lisen sleep lisen sleep Schedule 1 Schedule 2 sync Wei Ye, USC/ISI
B-MAC Componens Channel arbiraion Clear Channel Assessmen (CCA) back offs Reliabiliy Link layer acknowledgemens Power efficien communicaion Low Power Lisening (LPL) Noe: services like organizaion, synchronizaion, and rouing are lef o higher levels.
Clear Channel Assessmen Auomaic esimaion of noise floor Simple hreshold reduces hroughpu If no ouliers afer 5 samples, channel is considered busy
Clear Channel Assessmen Configurable knobs Enable/Disable CCA Configure iniial and congesion back off imes Adjuss proocol s Fairness Available hroughpu
Low Power Lisening (LPL) Higher level communicaion scheduling Energy Cos = RX + TX + Lisen Sar by minimizing he lisen cos Example of a ypical low level proocol mechanism Periodically wake up, sample channel, sleep Properies Wakeup ime fixed Check Time beween wakeups variable Preamble lengh maches wakeup inerval Overhear all daa packes in cell Duy cycle depends on number of neighbors and cell raffic Node 1 Node 2 wakeup sleep wakeup wakeup sleep RX sleep sleep wakeup TX sleep wakeup wakeup sleep wakeup ime ime
Effec of LPL Check Inerval Single hop daa reporing applicaion Higher sampling rae Higher raffic in a cell Higher duy cycle Opimize he check ime o he raffic Applicaion knows sample rae (packe generaion rae) Lifeime (years) 4 3.5 3 2.5 2 1.5 1 0.5 Effec of sample period on node duy cycle 1-min sample period 5-min sample period 10-min sample period 20-min sample period 0 0 50 100 150 200 Check Time (ms)
Implemenaion Size Higher level service buil on op of B-MAC in order o compare wih S- MAC Reliable ranspor (Acks) Hidden Terminal suppor (RTS-CTS) Implemenaion smaller han S-MAC
Fragmenaion Suppor Facored vs Layered Proocol S-MAC RTS-CTS Fragmenaion Suppor B-MAC Nework proocol sends iniial daa packe wih number of fragmens pending Disable backoff & LPL for res of fragmens Measure energy consumpion a C (boleneck node) Minimizing power relies on conrolling link layer primiives A B C E D Energy per bye (mj/bye) 0.4 0.35 0.3 0.25 0.2 0.15 0.1 Mean energy consumpion per bye (10 second sample period) 0.4 B-MAC w/ no app conrol B-MAC w/ app conrol S-MAC 0.35 T-MAC (simulaed) Opimal Schedule 0.3 Energy per bye (mj/bye) 0.25 0.2 0.15 0.1 Mean energy consumpion per bye (100 second sample period) B-MAC w/ no app conrol B-MAC w/ app conrol S-MAC T-MAC (simulaed) Opimal Schedule Someimes he black box is worse han he naïve approach 0.05 0.05 0 0 0 50 100 150 200 250 0 50 100 150 200 250 Fragmen size (byes) Fragmen size (byes)
Tradeoffs: Laency for Energy Facored vs Tradiional Proocol Energy (mj) 550 500 450 400 350 300 250 200 150 100 50 Effec of laency on mean energy consumpion 11 10 9 3 2 1 B-MAC S-MAC Always On S-MAC Defaul Configuraion 0 0 2000 4000 6000 8000 10000 Laency (ms) Assume a mulihop packe is generaed every 10 sec No queuing delay allowed Delay he packe S-MAC sleeps longer beween lisen period B-MAC increases he check inerval and preamble lengh B-MAC Defaul Configuraion
Tradeoffs: Throughpu for Energy Facored vs Layered Proocol 10 node single hop nework Increase ransmission rae Deliver each packe wihin 10 sec Measure average power consumpion per node As hroughpu increases B-MAC reduces check inerval as raffic increases S-MAC uses opimal duy cycle Proocol overhead causes energy o increase linearly Power consumed (mw = mj/second) 50 45 40 35 30 25 20 15 10 5 9 10 8 Effec of consan hroughpu on power consumpion B-MAC S-MAC Always On 1 7 2 opology 6 3 5 4 0 0 50 100 150 200 250 Throughpu (bis/second)
Lifeime Model min( E ) = Erx + Ex + Elisen + E Transmi x E x = r ( L + L ) = x c xb Receive rx E rx c preamble V V packe xb nr ( L + L ) = rx rxb preamble packe sleep rxb Noaion r n L preamble L packe c sleep c rxb c xb C ba V i sarup rxb rx xb x l Parameer Sample Rae (packes/sec) Neighborhood size Preamble lengh (byes) Packe lengh (byes) Curren : Sleep (ma) Curren : Rx one bye Curren : Tx one bye Capaciy : Baery (mah) Volage Time : Radio sampling inerval (s) Time : Radio sarup Time : Rx one bye Time : Rx per second Time : Tx one bye Time : Tx per second Time : Lifeime (s)
Lifeime Model min( E ) = E + E + E + rx LPL Sampling x lisen E sleep Noaion r Parameer Sample Rae (packes/sec) E E Sleep lisen sleep E sample lisen sleep = = 17.3µ J = E = 1 sarup sample rx sleep 1 1 c i i x sleep lisen n L preamble L packe c sleep c rxb c xb C ba V i sarup rxb rx xb x l Neighborhood size Preamble lengh (byes) Packe lengh (byes) Curren : Sleep (ma) Curren : Rx one bye Curren : Tx one bye Capaciy : Baery (mah) Volage Time : Radio sampling inerval (s) Time : Radio sarup Time : Rx one bye Time : Rx per second Time : Tx one bye Time : Tx per second Time : Lifeime (s)
Lifeime Model min( E ) = E + E + E + The oal energy, E, can be used o calculae he expeced lifeime of he sysem l = C rx ba E V x lisen 60 60 E sleep Noaion r n L preamble L packe c sleep c rxb c xb C ba V i sarup rxb rx xb x l Parameer Sample Rae (packes/sec) Neighborhood size Preamble lengh (byes) Packe lengh (byes) Curren : Sleep (ma) Curren : Rx one bye Curren : Tx one bye Capaciy : Baery (mah) Volage Time : Radio sampling inerval (s) Time : Radio sarup Time : Rx one bye Time : Rx per second Time : Tx one bye Time : Tx per second Time : Lifeime (s)
2.5 2 Effec of Neighborhood Size Neighborhood Size affecs amoun of raffic in a cell Nework proocols ypically keep rack of neighborhood size Bigger Neighborhood More raffic Channel Aciviy Check Inerval (ms) 200 180 160 140 120 100 80 60 40 1.5 1.25 1 Expeced Lifeime Conour 0.75 0.5 0.25 Effecive duy cycle (%) 0.1 0.09 0.08 0.07 0.06 0.05 0.04 0.03 0.02 Effec of neighborhood size on node duy cycle 200ms check inerval 100ms check inerval 50ms check inerval 25ms check inerval 10ms check inerval 20 0.01 0 0 20 40 60 80 100 Neighborhood size 0 0 20 40 60 80 100 Number of neighboring nodes
Conclusions Coordinaion wih higher proocols is essenial for long lived operaion Tradiional absracion a he nework layer doesn fi sensor neworks need a new absracion a he link layer like B-MAC