1. Wireless Media Access Protocols Hari Balakrishnan LCS and EECS Massachusetts Institute of Technology http://www.sds.lcs.mit.edu/~hari hari@lcs.mit.edu

2. The Problem Shared communication medium Need to resolve contention amongst contending hosts o Efficiency: want high channel utilization o Fairness: want equitable allocation o “Distributedness”: preferable, but sometimes not done In fact, any shared medium (e.g., Ethernet) has this problem Wireless makes it a lot worse

3. Wireless Media Access: Problems Asymmetric communication properties Hidden terminals: A and B can talk; B and C can talk but A and C can’t Exposed terminals: B sending to A prevents C from sending because it senses a transmission in progress Collision detection is not possible And carrier sense has limitations (but is in fact done in many commercial products)

4. CSMA protocols Stands for Carrier Sense Multiple Access CSMA/CD & CSMA/CA popular varieties CSMA/CA for wireless o Before transmission, sense carrier o If not busy, go ahead and transmit (sometimes after a small time gap). o If busy, defer transmission (“backoff”), then try again as before o Successive backoffs are exponential o E.g., WaveLAN and other commercial products

5. Problems with CSMA/CA CA heuristics Small implementation bugs tend to distort performance heavily Doesn’t solve the hidden/exposed terminal problems CSMA tries to detect contention near a sender, when in fact it’s the receiver we should worry about!

6. MACA: Control Internet PT ER FH GW MH Modem PR ER PT Fixed Host Ethernet Radios Poletop Radios Mobile Host RTS CTS Want synchronization before communication

7. Data MACA: Data Internet PT ER FH GW MH Modem PR ER PT Fixed Host Ethernet Radios Poletop Radios Mobile Host Data Ack RTS No response Exponential backoff

8. MACAW Enhances MACA and tunes it to work in a “nano-cellular” wireless environment Key innovations o When station transmits, send current backoff value; all stations pick this up o Multiplicative increase, linear decrease o Adds an ACK to MACA o DS message with length (no carrier sense hardware) Result: RTS-CTS-DS-DATA-ACK

9. Dense Packet Radio Networks Motivation o Can we ignore aggregate interference from stations far away? o Can we come up with a robust scheduling scheme? Spread-spectrum o Direct sequence: spreads signal in frequency o Frequency-hopping: keep changing frequency o CDMA: orthogonal codes for stations o (Interference looks like random noise)

10. Do Stations Far Away Affect Us? Cute, back-of-the-envelope analysis based on physics (Show on board) Result: S/N goes as 1/(ln M) Assumes a two-dimensional world! Power-controlled transmissions improve scaling

11. Minimum-Energy Routing D d d More energy-efficient to go through B than directly But it adds latency

12. Scheduling Publish random schedules (Similar to random frequency hopping sequences in FH spread spectrum) Q: Paper says receiver duty cycle of 0.3 is good. ???

13. Summary CSMA has limitations (but does have value) MACA/MACAW especially useful in packet radio o RTS-CTS exchange part of 802.11 spec Need a good way to evaluate fairness o Problem much harder than in wired settings Shepard’s paper shows that there’s no physical reason large-scale wireless can’t happen Randomized schedules for distributed and scalable access