1. Deconstructing Interference relations in 802.11 Networks Guided by Submitted by Prof. K. Chebrolu Akshat Saxena (08305016) MTP Stage I CS 697

2. Motivation Many cities have come up with WMN deployments ◦ Eg: MIT Roofnet, Strik’s Access/One ◦ Directly affects the throughput ◦ Important factor in routing, load and Call Admission Control

3. Definition – Interference Map Interference map is A collection of data structures ◦ Sending interference (Carrier Sense) ◦ Receiving interference (Hidden Terminal) Characterize the interaction among communicating nodes Helps in routing, load balancing and call admission control

4. Interference Basics Communication Range Packets sent by B can be received at A The actual distance depends on conditions, carrier frequency, and output power A B

5. Interference Basics Carrier Sense Range Includes communication range C’s carrier can be sensed at A and hence A backs off if C has transmission of its own Both these ranges can be Asymmetrical C A

6. Interference Basics Interference Range Assume B and D to be out of CS range of each other It is defined as all positions of D that destroy some of the B’s packets at A If (power(B A )- power(D A )) exceeds capture_threshold then safe else B’s packets are lost D A B

7. External Range None of E’s packets can destroy B’s packets at A But such nodes can together generates enough power to disrupt the communication Interference Basics E D C A B

8. Interference Basics (contd …) Sending InterferenceReceiving Interference Some nodes may reduce sender’s sending capacity by being in its CS range Interference occurring at receiver which the sender is unaware of What can be sent on airWhat can be received Non-destructiveDestructive S34 => 1 means 100% of times 3 can sense 4 S34 => 0.6 means 60% of times 3 can sense 4 For any receiver k, d34 => 1 means 4 is out of CS range of 3 d34 => 0.5 means either 3,4 share medium or 4 interferes with k

9. Why is it not trivial? Interferer’s identity can not be determined ◦ Cannot decode packets received from node present in Interference Range Topology of networks may change over time ◦ Characteristic feature of wireless It’s not a binary phenomenon but a fuzzy one Wi-fi devices are not the only interferers ◦ Devices like Bluetooth, zigbee, microwave ovens, cord-less phones and many more can affect communication

10. Potential Interferers Bluetooth DevicesMicrowave ovenZigbee sensors Motes Wireless devicesCordless phones

11. Related work

12. Approach category Interference map Measurement Based Link Interference Link Interference ratio Broadcast Interference ratio SNR / SINR Energy Sensing ESense

13. Approach category Interference map Measurement Based Link Interference Link Interference ratio Broadcast Interference ratio SNR / SINR Energy Sensing ESense

14. Link Interference Ratio (LIR) [JS05] IK

15. J IK

16. In any N node network, ◦ Computation time is O(n 2 ) while storage requirement is O(n 3 ) ◦ denotes the receiving interference What if more than 3 nodes are involved ? AdvantageDisadvantage Never miss any interfererSpace requirement is high Easy to implementNot scalable for large networks Need network downtime Link Interference Ratio (LIR) [JS05]

17. Broadcast Interference Ratio (BIR) [JS05] Instead of unicast, nodes now broadcast In one broadcast, all nodes in its Carrier sensing range are updated AdvantageDisadvantage Number of experiments reduced to O(n 2 ) Space requirement is high Easy to implementIf a broadcast packet is lost, it’s not retransmitted Need significant downtime

18. SIR based Interference mapping scheme [RJ09] Instead of delivery ratios, they calculate SIR at receiver – RSSI S – RSSI I Mainly concentrated on TDMA based wireless network AdvantageDisadvantage Number of experiments reduced to O(n) Captures only receiver side interference Easy to implementNeed significant downtime

19. Esense – Energy Sensing [KC09] An interference detection framework ◦ Detects interferer which is outside communication range but within carrier sensing range ◦ Senses energy for detection ◦ Embeds the node’s identity into packet length duration  Eg: For id = 5, choose packet of length 500  Eg: For id = 15, choose packet of length 1500

20. Esense – Energy Sensing Pre-built vocabulary needed at both sender and receiver If AP1can sense the ID of any other node, then that node is a potential interferer of AP1 Source: [KC09]

21. Approaches at a glance ApproachesLink interference Broadcast interference SIR basedESense Parameters MethodologyUnicast, LIRBroadcast, BIRSIRBroadcasted esense id packets What signifies presence of interferer 0 < LIR < 10 < BIR < 1SIR exceeding some threshold If u can sense esense ID of any node Network Downtime Needed Not needed Number of experiments needed O(n 4 )O(n 2 )O(n)Inline with ongoing communication Space OverheadO(n 3 ) O(n)Vocabulary lookup Detection RangeTill Interference Range Till Interference Range of receiver Till Carrier sensing Range Additional Overhead - -Sensing hardware

22. Deconstructing Interference Relations for 802.11 networks AIM To build an interference map for wireless networks which can capture both sender and receiver side interference First Step Goal To perform controlled experiments to understand the behavior of both sender and receiver side interference Road Ahead Use our understanding from first step on Esense framework to design our interference map

23. The First Step Card Calibration Controlled sending interference Controlled receiving interference

24. Image courtesy – www.titanwireless,com Card Calibration Aim ◦ To find out if wireless cards adhere to specifications Experimental Setup Ubiquity mini-pci SR2 Agilent CSA Spectrum Analyzer N1996A-503 Laptop with a wifi card XR2 cardSR5 card SRC card LMR 400 RF cable

25. Card Calibration For each 1dbm increase in transmitted power, different cards show different increase. Avoid using these cards in range < 5dbm

26. Sending Interference AIM ◦ To understand the effect of sending interference in controlled settings ◦ The threshold beyond which the effect of interferer is negligible ◦ Does the throughput increases drastically from 0 to 1 or in steps

27. Sending Interference Sender (S) Receiver (R) Interferer (I) 50db Attenuator 50-90db Attenuator Experimental Setup Sender config: CS not disabled Back off on/off Continuous Tx RSSI at S due to I Interferer config: CS disabled Back off disabled Continuous Tx Receiver: Throughput at R from S

28. Sending Interference

29. Sender beginning to get stronger Interferer stronger Interferer almost like inactive

30. Receiving Interference AIM ◦ To understand the effect of receiving interference in controlled settings ◦ The threshold beyond which the effect of interferer is negligible ◦ Does the throughput increases drastically from 0 to 1 or in steps

31. Receiving Interference Experimental Setup Sender config: CS not disabled Back off on Continuous Tx Sender config: CS disabled Back off disabled Continuous Tx

32. Receiving Interference

33. Interferer is stronger in this region If this is the difference in their RSSIs, then no loss due to interference

34. The Road Ahead … Setup a testbed for Esense Use the knowledge from the calibration experiments to design a methodology for constructing an interference map

35. Acknowledgement I would like to thank Prof. K. Chebrolu for her help and encouragement at each step of this study. Sincere thanks to Prof. Girish Kumar, faculty (EE), IIT Bombay for allowing us to perform experiments in his lab. Special thanks to Prof. P. Kulkarni for his valuable suggestions and time during calibration experiments. Thanks to Anup Kunte,Vishal Sevani and other students of CSE, IIT Bombay, for their help and support during the experiments.

36. References [DN07] “ Interference Map for 802.11 Networks”, Dragos Niculescu, Internet Measurement Conference (IMC) 2007 [JS05] “Estimation of Link Interference in Static Multi-hop Wireless Networks”, Jitendra Padhye, Sharad Agarwal, Venkata N. Padmanabhan, Lili Qiu, Ananth Rao, Brian Zill, Internet Measurement Conference (IMC) 2005 [CR06]“Measurement-Based Models of Delivery and Interference in Static Wireless Networks”, Charles Reis, Ratul Mahajan, Maya Rodrig, David Wetherall, John Zahorjan, ACM SIGCOMM 2006 [SD06]“Characterizing MultiWay Interference In Wireless Mesh Networks”, Saumitra M. Das, Dimitrios Koutsonikolas, Y. Charlie Hu, and Dimitrios Peroulis, Wintech 2006 [KC09] “ Esense: Communication through Energy Sensing”, Kameswari Chebrolu, Ashutosh Dekhne, ACM Mobicom 2009 [RJ09] “Evaluation of the Signal to Interference Ratio based Interference Mapping Strategy on a 802.11b/g based Wireless Mesh Network”, Rahul Jain, Mtech thesis under Prof. Bhaskaran Raman

37. Thank You Any Questions !

38. Mote – Wifi Range Experiment Aim ◦ If sender(interferer) present beyond interference range can be identified Setup Sender laptop with mote Sender laptop Esense packet Distance varied

39. Mote – Wifi Range Experiment

40. It is possible to detect a sender present beyond interference range Wifi can’t listen but mote can