Outline Medium ccess ontrol With oordinated daptive Sleeping for Wireless Sensor Networks Presented by: rik rooks Introduction to M S-M Overview S-M Evaluation ritique omparison to MW Washington University epartment of omputer Science and Engineering M (Medium ccess ontrol) ll shared-medium networks need an effective M protocol ontrols access to the shared medium Ensures no two nodes interfere with each other s transmissions sends a message receives??? sends a message General M Requirements Ensure reliable communication Ensure fair access to shared physical transmission medium for all contending streams Minimize delay for sending/receiving messages Maximize bandwidth utilization Typical M protocols ontention-based M (arrier Sense Multiple ccess) M/MW (Virtual Sensing RTS/) IEEE 802.11 (Physical & Virtual Sensing) PMS -based TM (Time-ivision Multiple ccess) Others M (ode-ivision Multiple ccess) Typical M protocols ontention-based Pros voids collisions well, no synchronization ons ontrol overhead, idle listening -based Pros onserves energy ons oesn t scale well, synchronization difficult 1
M Requirements for Wireless Sensor Networks Energy Efficiency Scalability/daptability to network size, density, and topology Less emphasis on fairness, latency, throughput, and bandwidth utilization S-M: esign Uses elements of contention-based and schedule-based M protocols Tries to reduce energy waste from all major sources ollisions, Overhearing, ontrol packet overhead, Idle Listening Trades performance for energy efficiency Increases per-hop delay Reduces fairness s of Energy Waste ollisions Re-transmitting packets takes a lot of energy Overhearing Receiving packets destined to other nodes ontrol packet overhead Sending/Receiving control packets Idle Listening n idle radio consumes almost as much power as during reception S-M: Low uty-ycle Operation Low duty-cycle operation Reduces idle-listening by putting nodes to sleep most of the time However, increases latency, which can accumulate on each hop Listen Sleep Listen Sleep L S-M: oordinated Sleeping Uses oordinated Sleeping to reduce control overhead and latency Nodes adopt their neighbors sleep schedule If conflicting schedules, nodes can adopt both or just remember the second for transmit Only initiate communication during awake time, during which neighbors should also be awake Periodically broadcast SYN message to maintain sync Periodic neighbor discovery (listens to whole sync period) S-M: Listen Listen broken into slots for SYN, RTS, Receiver Listen SYN Transmitter Tx SYN Tx RTS Rx Rx ata/sleep 2
S-M: ollision voidance ollision and Overhearing voidance Uses physical N virtual carrier sense RTS//ata/ck ll nodes that hear a RTS or sleep for the specified TX time to avoid overhearing/collision Inspired by PMS Sender Receiver E F S-M: Message Passing Sending Long Messages is costly in terms of energy and latency Re-transmission of a long message is costly Transmitting many small fragments requires extra overhead Message Passing used to minimize costs of sending a long message Use a single RTS/ for all fragments, reducing control packet overhead Re-transmission limited to corrupted fragments S-M: daptive Listening Uses daptive Listening to minimize delay Nodes that hear RTS/ wake up and listen for a short time immediately after the previous transmission should have ended Not as good as it sounds because RTS/ packets exchanged during adaptive wakeup period are not during the next packets wake time S-M: Example w/o daptive Listening K Rx ata dded elay S-M: Example with daptive Listening Rx K daptive Not Seen Rx K daptive RTS Not Seen dded elay S-M: Energy Performance Two-hop network Savings due to avoiding overhearing and efficiently transmitting long messages 1 2 2 E 1 3
S-M: Energy Performance S-M: Latency Performance Lowest Traffic Load Highest Traffic Load S-M: Throughput Performance S-M: Energy-Time ost Performance 1 2 3 10 11 Highest Traffic Load nother Traffic Load Key Ideas ombines advantages of contention-based (good collision avoidance) and schedulebased (energy efficiency) Low-duty cycle with coordinated sleeping RTS//ata/ck when awake Overhearing avoidance based on info in RTS/ packets daptive sleeping used to reduce latency compared to other schedule-based protocols ritique/questions The authors dismiss fairness as an important part of the M protocol because of an unsubstantiated assumption about the nature of wireless sensor network applications How does S-M compare to other M protocols designed for energy efficiency (TM, etc )? o schedules ever combine or get reformed to synchronize more groups of nodes? This doesn t seem to meet the requirement of being flexible to network changes 4
omparison to MW Where to learn more MW esigned to improve fairness, throughput, latency, and reliability RTS--S-T-K ynamic backoff algorithm (lower if contention level is low) S-M esigned to improve energy efficiency dds coordinated sleep for energy conservation ackoff algorithm is random sleep Uses fragmentation to reduce control overhead http://www.isi.edu/scadds/projects/smac/ General info and links related to S-M http://www.isi.edu/ilense/software/smac/ ownload source code FQ http://www.isi.edu/~weiye/pub/commstack.pdf etailed description of the modified comm stack used to implement S-M on TinyOS 5