1. TAPs: An Architecture and Protocols for a High-Performance Multi-hop Wireless Infrastructure Ed Knightly ECE/CS Departments Rice University http://www.ece.rice.edu/~knightly Joint work with V. Kanodia, A. Sabharwal, and B. Sadeghi

2. Ed Knightly The Killer App is the Service l High bandwidth l High availability –Large-scale deployment –High reliability –Nomadicity l Economic viability l Why? –Broadband to the home and public places –Enable new applications

3. Ed Knightly WiFi Hot Spots? l Why? poor economics –High costs of wired infrastructure ($10k + $500/month) –Pricing: U.S. $3 for 15 minutes; CH: 0.90 CHF/minute –Dismal coverage averaging 0.6 km 2 per 50 metro areas projected by 2005 l 11 Mb/sec, free spectrum, inexpensive APs/NICs Carrier’s Backbone/Internet T1 Medium bandwidth (wire), sparse, and expensive

4. Ed Knightly Cellular? l Cellular towers are indeed ubiquitous –Coverage, mobility, … l High bandwidth is elusive –Aggregate bandwidths in Mb/sec range, per-user bandwidths at dial-up speeds –Expensive: spectral fees and high infrastructure costs High availability, but slow and expensive

5. Ed Knightly Ad Hoc Networks? l Availability –Problems: intermediate nodes can move, power off, fade, DoS attack  routes break, packets are dropped, TCP collapses, … l Low bandwidth –Poor capacity scaling –Unlike cellular, users consume wireless resources at remote locations “Free” but low availability and low bandwidth

6. Ed Knightly TAPs: Multihop Wireless Infrastructure l Transit Access Points (TAPs) are APs with –beam forming antennas –multiple air interfaces –enhanced MAC/scheduling/routing protocols l Form wireless backbone with limited wired gateways

7. Ed Knightly Multihop Wireless Infrastructure l Transit Access Points (TAPs) are APs with –beam forming antennas –multiple air interfaces –enhanced MAC/scheduling/routing protocols l Form wireless backbone with limited wired gateways l High bandwidth –High spatial reuse and capacity scaling –Opportunistic protocols l High availability –Redundant paths and non-mobile infrastructure –Deployability l Good economics –Unlicensed spectrum, few wires, exploit WiFi components

8. Ed Knightly Prototype and Testbed Deployment l FPGA implementation of enhanced opportunistic, beamforming, multi-channel, QoS MAC l Build prototypes and deploy on Rice campus and nearby neighborhoods l Measurement study from channel conditions to traffic patterns

9. Ed Knightly Outline l TAP architecture l OAR: an opportunistic auto-rate MAC l MOAR: multi-channel OAR l Open problems

10. Ed Knightly l Received signal: superposition of different reflections, with different delays and attenuations Motivation l Wireless channel is variable l Coherence time channel gain  time 

11. Ed Knightly Opportunistic MAC Goal l Constraint: distributed random access protocol l Exploit the variations inherent in wireless channel to increase throughput l Maintain fair temporal shares for different flows channel gain  time  user 1 user 2 user 1 user 2

12. Ed Knightly IEEE 802.11 Multi-rate l Support of higher transmission rates in better channel condition –802.11b available rates: 2, 5.5, 11 Mbps –802.11a available rates: up to 54 Mbps l Auto Rate Fallback (ARF) –[Monteban et al. 97] –Use history of previous transmissions to adaptively select future rates

13. Ed Knightly Temporal vs. Throughput Fairness l Equivalent in single-rate networks l Throughput fairness results in significant inefficiency in multi-rate networks Example user 1 user 2 access point user 3

14. Ed Knightly Temporal vs. Throughput Fairness l Equivalent in single-rate networks l Throughput fairness results in significant inefficiency in multi-rate networks Example user 1 user 2 access point user 3 Throughput Fair user 1 user 2 DATA user 3 DATA Even 1 user with low transmission rate results in a very low network throughput

15. Ed Knightly Temporal vs. Throughput Fairness l Equivalent in single-rate networks l Throughput fairness results in significant inefficiency in multi-rate networks Example user 1 user 2 access point user 3 Temporal Fair user 1 user 2 DATA user 3 DATA Same time-shares of the channel for different flows, also higher throughput

16. Ed Knightly Opportunistic MAC Goal l Exploit short-time-scale variations inherent in wireless channel to increase throughput in wireless ad hoc networks Issue l Maintaining temporal share of each node Challenge l Channel info available only upon transmission

17. Ed Knightly Opportunistic Auto Rate (OAR) l Observation: coherence time on order of multiple packet transmission times –measure channel quality on RTS/CTS handshake –hold good channels for multiple transmissions l Ensure fairness by scaling number of packets transmitted to channel quality –# packets = Current rate / Base rate –with random access, all flows equally likely to access channel OAR: High throughput, while maintaining temporal fairness properties of single rate IEEE 802.11

18. Ed Knightly RTS RBAR Protocol source destination ACK CTS Receiver Based AutoRate (RBAR) DATA l Receiver controls the sender’s transmission rate l Control messages sent at Base Rate Reservation Sub-Header

19. Ed Knightly DATA RTS ACK CTS ACK OAR Protocol source destination OAR l Once access granted, it is possible to send multiple packets if the channel is good Reservation Sub-Header

20. Ed Knightly Observation I Time spent in contention per packet is the same for RBAR and single-rate IEEE802.11 Transmitter Receiver OAR Performance Comparison IEEE 802.11 R CA D1 Transmitter Receiver R CA D1R CA Transmitter Receiver RBAR Observation II OAR contends for the same total time as singe-rate IEEE 802.11 but transmits more data R CA D1 A D2 A D3

21. Ed Knightly Transmitter Receiver OAR Performance Comparison IEEE 802.11 R CA D1 Transmitter Receiver R CA D1R CA Transmitter Receiver RBAR Observation III OAR holds high-quality channels for multiple transmissions R CA D1 A D2 A D3

22. Ed Knightly Analytical Model l Challenge: MAC and channel are random processes with memory l Model relates physical-layer characteristics to MAC throughput: –Time spent in contention  Markov model of modified IEEE 802.11 –Average transmission rate  Due to channel distribution l Comparative model b/t multi-rate OAR and tractable systems –TIME: OAR contends as often as single-rate IEEE 802.11 with increased data per contention –PACKETS: OAR reduces packet transmission time via per- contention rate adaptation

23. Ed Knightly Simulation Results Under Ricean Fading l OAR has 42% to 56% gain over RBAR l Increase in gain as number of flows increases l Model predicts OAR & RBAR throughput to within 7% accuracy Nodes

24. Ed Knightly Outline l TAP architecture l OAR: an opportunistic auto-rate MAC l MOAR: multi-channel OAR l Open problems

25. Ed Knightly –Example: at 2.4 GHz WiFi, 5 vs. 1-3 MHz –Figure for Ricean, K=4 Multi-Channel Problem Formulation l Observe: for two MUs, quality of different channels can have low correlation if channel separation >> coherence bandwidth

26. Ed Knightly Challenge l Ideal protocol is simple: select the best channel at the instant of transmission l In practice, channel qualities are unknown a priori –Must first transmit and measure l Cost of measuring channels must be balanced with benefits of finding good ones

27. Ed Knightly MOAR Protocol Sketch l Measure channel SNR at RTS/CTS handshake l If channel quality is high (above an SNR threshold), transmit via OAR l If channel quality is poor, skip to a new channel –next channel piggybacked in CTS l Design optimal stopping rule for skipping –stop when throughput gain of skipping to a better channel is outweighed by overhead l Ensure fairness

28. Ed Knightly Optimal Stopping Rule Formulation l Let X n denote the SNR of the n th measured channel l Let c denote the cost (in time) of measuring the channel l After observing X n transmit or measure again? –cannot go back to previous channel (coherence time) l The reward for the n th selection is X n -nc –after scaling SNR to rate and then to time l Objective: maximize the expected reward l In a class of stopping rule problems (without recall)

29. Ed Knightly Optimal Stopping Time l Let V* denote the expected return from the optimal stopping rule l Suppose pay c and observe X 1 = x 1 l If continue, x 1 is lost and c is paid –continuing, can obtain return V*, but not more –start afresh l Optimal rule is threshold based –If x n V* stop –N* = min{n  1: X n  V*}

30. Ed Knightly Calculating the Stopping Threshold V* l V* = E max(X 1,V*) – c l –F(x) represents the SNR distribution l Compute V* –channel model and parameters (ex. K, d) –system’s rate-SNR thresholds (ex. 1, 2, 5.5, 11)

31. Ed Knightly MOAR Throughput Gains Ricean parameter K = 0 is no line-of-sight signal l Gains of 40%-60% increasing with K and SNR variance

32. Ed Knightly Effect of Node Distance l Greatest help when far away l Non-monotonic due to rate-SNR thresholds

33. Ed Knightly Random Topologies l Nodes are uniform-randomly placed in a 250m circle l “Optimal Skipping” cheats: looks at all channels (with no cost) and jumps to the best l Observe –MOAR extracts most available gain –close-by nodes detract from average gain

34. Ed Knightly Outline l TAP architecture l OAR: an opportunistic auto-rate MAC l MOAR: multi-channel OAR l Open problems

35. Ed Knightly DoS Resilience and Security l Old methodology –Design a network protocol –Optimize for performance –Discover DoS/Security holes  Ex. Route query floods –Patch one-by-one l Challenge –DoS-resilience and security as the foundation of network protocols –Recognize these issues are as important as performance

36. Ed Knightly TAP Media Access and Scheduling l Challenge: distributed scheduling –Others’ channel states, priority, & backlog condition unknown  Ex. TAP A’s best recv’r may be transmitting elsewhere  Ex. Traffic to be recv’d may be higher priority than that to be sent –Traffic and system dynamics preclude scheduled cycles –Modulate aggressiveness according to overheard information

37. Ed Knightly Multi-Destination Routing/Scheduling l Most data sources or sinks at a wire l Routing protocols for any wire abstraction l Scheduling –At fast time scales, which path is best (channels, contention, …) now? –Can delay/throughput gains be realized despite TCP?

38. Ed Knightly Distributed Traffic Control l Distributed resource management: how to throttle flows to their system-wide fair rate? –Throttle traffic “near-the-wire” to ensure fairness and high spatial reuse –TCP cannot achieve it (too slow and RTT biased) –Incorporate channel conditions as well as traffic demands

39. Ed Knightly Capacity Driven Protocol Design Protocol Driven Capacity Analysis l Traditional view of network capacity assumes zero protocol overhead (no routing overhead, contention, etc.) l Protocols themselves require capacity l A new holistic system view: “the network is the channel” –Incorporate overhead in discovering/measuring the resource –Explore capacity limits under real-world protocols

40. Ed Knightly Problem: Multiple APs/TAPs/… within Radio Range l PHY Interference has disproportionate throughput degradation at MAC layer l Interference can lead to severe scaling limitations and starvation (worse than zero-sum game)

41. Ed Knightly Summary l Transit Access Points –WiFi “footprint” is dismal –Removing wires is the key for economic viability l Opportunistic Scheduling (OAR/MOAR) –Exploit time and frequency diversity l Challenges –Multi-hop wireless architectures –Distributed control –Scalable protocols