2. Fair Sharing of MAC under TCP in Wireless Ad Hoc Networks Mario Gerla Computer Science Department University of California, Los Angeles Los Angeles, CA 90095 http://www.cs.ucla.edu/NRL/wireless

3. UCLA DARPA Domains Project Outline Overview of CSMA, FAMA and IEEE 802.11. MAC performance with TCP.  Variable Hop Length Experiments.  Hidden Terminal Experiments.  Ring Experiments.  Grid Experiments. Static. Mobility.

4. UCLA DARPA Domains Project Simulation Using GloMoSim Detailed model of the protocol stack. Allows investigation of TCP and MAC layer interactions. Capability to simulate large number of nodes. GloMoSim web page.  http://pcl.cs.ucla.edu/projects/domains/glomosim.html

5. UCLA DARPA Domains Project MAC Layer Protocols CSMA  Requires carrier sensing before transmission.  If the channel is free, the packet is transmitted immediately.  Otherwise, it is rescheduled after a random timeout. FAMA  Builds on CSMA.  Uses the RTS (Request To Send) and CTS (Clear To Send) exchange to prepare the floor for data transmission. 802.11  Uses carrier sensing and RTS/CTS, similar to FAMA.  Utilizes link-level ACKs.  Collision Avoidance scheme.

6. UCLA DARPA Domains Project Variable Hop Length Experiments Configuration 012543 Each node is 10 meters apart from its neighbors. Each node has a radio power range of 10 meters. 2Mbps channel bandwidth. FTP traffic. TCP window size varies from 1 to 16 packet size. Variable number of hops (single connection).  i.e., FTP connection 0-1, 0-2, 0-3, 0-4, 0-5 (one at a time).

7. UCLA DARPA Domains Project Variable Hop Length Experiments Results

8. UCLA DARPA Domains Project Variable Hop Length Experiments Results (Cont’d)

9. UCLA DARPA Domains Project Variable Hop Length Experiments Results (Cont’d) CSMA and FAMA degrades with window size > 1 pkt.  Collisions between TCP data and ACKs. 802.11 performs the same no matter the window size.  Link-level ACKs combat collisions.

10. UCLA DARPA Domains Project Hidden Terminal Experiments Configuration FTP traffic Connections from node 0 to node 1 and from node 2 to node 1. Node 0 and node 2 cannot hear each other. 0 1 2

11. UCLA DARPA Domains Project Hidden Terminal Experiments Results CSMA suffers from hidden terminal. FAMA and 802.11 performs well due to RTS/CTS exchange.

12. UCLA DARPA Domains Project Ring Experiment Configuration FTP traffic Multiple single hop connections.  i.e., 0-1, 1-2, 2-3, 3-4, 4-5, 5-6, 6-7, 7-0 (at the same time). 0 1 6 5 4 3 2 7

13. UCLA DARPA Domains Project Ring Experiment Results FAMA best in both fairness and aggregate throughput. 802.11 unfairness due to timers.

14. UCLA DARPA Domains Project 802.11 Fairness Yield time.  Time a node yields after transmitting a frame.

15. UCLA DARPA Domains Project Grid Experiment Configuration Each node is 10 meters apart from its horizontal and vertical neighbors. Each node has a radio power range of 30 meters. FTP connections are established between node 18 to node 26, node 36 to node 44, node 54 to node 62, node 2 to node 74, node 4 to node 76 and node 6 to node 78. 0 1 10 9 8 2 19 18 17 11 20 74 73 72 26 80 3 4 13 12 5 22 21 14 23 77 76 75 6 15 7 24 16 25 79 78 27 28 37 36 35 29 44 38 30 31 40 39 32 41 33 42 34 43 46 45 47 53 49 48 50 51 52 54 55 64 63 62 56 71 65 57 58 67 66 59 68 60 69 61 70

16. UCLA DARPA Domains Project Grid Experiment Configuration (Cont’d) 2Mbps channel bandwidth. Nodes move at a rate of 10 meters per second in a random direction with a probability of 0.5. When mobility is not considered, static routing is used. When mobility is introduced, Bellman-Ford routing is utilized with routing table updates occurring once every second.

17. UCLA DARPA Domains Project Grid Experiments Results (No Mobility) Without mobility  CSMA performs poorly due to interference by neighboring streams and by intersecting streams.  FAMA fair due to RTS/CTS and less aggressive yield time.  802.11 exhibits capture.

18. UCLA DARPA Domains Project Grid Experiments Results (With Mobility)  CSMA and FAMA collapse with mobility due to lack of fast loss recovery facilities.  802.11 still operational. Link level ACKs help recover from loss caused by transient nodes. Capture exists.

19. UCLA DARPA Domains Project Bluetooth Experiments Gerla, M et al, Tyrrenia Conf, sept 2000 Experiment #!: TCP throughput in a single piconet. Throughput versus the no. of TCP connections. Each TCP connection starts from a different slave on the common piconet, and goes through the access point (BT master). Experiment #2: TCP throughput when multiple piconets are used in parallel. Each piconet here supports a separate TCP connection. Experiment #3: TCP and IP Telephony in a multiple piconet configuration. IP Telephony uses ACL channel. Question: can TCP and Telephony coexist?

20. UCLA DARPA Domains Project S LAN IP backbone M 1 IP router M 2 M 3 LAN IP backbone IP router M 1 M 2 M 3 S 1 S 2 S 3 M 3 IP backbone M 2 M 1 LAN IP router S (a)(b)(c) Fig. 4.

21. UCLA DARPA Domains Project Exp # 1:TCP throughput in Bluetooth (single piconet)

22. UCLA DARPA Domains Project Exp #2:TCP Throughput in WaveLAN vs BT (multiple piconets)

23. UCLA DARPA Domains Project Exp #2 (cont): Throughput of TCP flows

24. UCLA DARPA Domains Project TCP and IP Telephony Voice carried on the ACL channel (ie, VoIP) Four piconets In each piconet: 1 TCP and 6 Voice connections TCP connections “always on” (file transfers) Voice: ON-OFF model; 8Kbps coding rate Voice packets: 20ms packetization -> 20 bytes With header overhead: voice pkt = 30 bytes

25. UCLA DARPA Domains Project Exp #3: Bluetooth; TCP + VoIP

27. UCLA DARPA Domains Project Exp #3: WaveLan : TCP + VoIP

28. UCLA DARPA Domains Project Exp #3: WaveLan : TCP + VoIP With 750 ms playout buffer, still 5% packets lost!

29. UCLA DARPA Domains Project Simulation: what have we learned? Fair sharing in BT across TCP connections (IEEE 802.11 is unfair, “capture”- prone) BT aggregate throughput exceeds IEEE 802.11 BT supports voice well even in heavy TCP load (IEEE 802.11 cannot deliver voice with TCP load)