1. Sep 12, 2002Wireless Networking Seminar1 Does IEEE 802.11 Work Well in Multi-hop Wireless Network? Author: Shugong Xu, Tarek Saadawi City University of New York Speaker: Weisheng Si

2. Sep 12, 2002Wireless Networking Seminar2 Outline Overview of The Paper Review of TCP Simulation Environment Instability Problem Unfairness Problem Summary of The Paper Comments on The Paper

3. Sep 12, 2002Wireless Networking Seminar3 Overview of The Paper Conclusion: Current 802.11 Protocol does not function well in multi-hop networks. Experiment Methods: –A Static String Topology –TCP as Transport Layer Protocol Evidences: –Instability Problem –Unfairness Problem 01234567

4. Sep 12, 2002Wireless Networking Seminar4 Review of TCP -- Sliding Window Goal: reliable data transmission Basic idea: –Using acknowledgement/retransmission scheme –Identifying data with sequence numbers Sliding Window: the maximum range of data sent but not acknowledged

5. Sep 12, 2002Wireless Networking Seminar5 Review of TCP -- Sliding Window [ 0 1 2 3 ] 4 5 6 7 8 9 0 1 2 3 ACK 1 0 [ 1 2 3 4 ] 5 6 7 8 9 4 1 2 3 4 An Example: Sliding Window Size = 4 bytes Timeout

6. Sep 12, 2002Wireless Networking Seminar6 Review of TCP—Flow Control Goal: prevent the buffer at the receiver from being overloaded. Basic Idea: –The receiver advertises its available buffer size to the sender in each TCP acknowledgment. –The sender maintains a variable called receiver window which specifies the size of the sliding window. Whenever it receives an acknowledgement from the receiver, it set the receiver window to the available buffer size advertised by the receiver.

7. Sep 12, 2002Wireless Networking Seminar7 Review of TCP – Flow Control [ 0 1 2 3 ] 4 5 6 7 8 9 0 1 2 3 ACK 1, WIN=4 0 [ 1 2 3 4 ] 5 6 7 8 9 4 ACK 5, WIN=3 0 1 2 3 4 [ 5 6 7 ] 8 9 5 6 7 An Example: Initial Sliding Window Size = 4 bytes

8. Sep 12, 2002Wireless Networking Seminar8 Review of TCP—Congestion Control Goal: prevent the network from being overloaded. Basic Ideas: –Using timeout as the indication of network congestion. –Additive Increase and Multiplicative Decrease. Two Additional Variables –Congestion Window –Slow Start Threshold

9. Sep 12, 2002Wireless Networking Seminar9 Review of TCP—Congestion Control Slow Start Threshold Slow Start Congestion Avoidance TCP Tahoe, Reno

10. Sep 12, 2002Wireless Networking Seminar10 Review of TCP – Congestion Control [ 0 1 2 3 ] 4 5 6 7 8 9 0 1 2 3 ACK 2, WIN=4 0 1 [ 2 3 4 5 ] 6 7 8 9 4 5 0 1 [ 2 ] 3 4 5 6 7 8 9 2 An Example: Sliding Window Size = 4 bytes Timeout

11. Sep 12, 2002Wireless Networking Seminar11 Review of TCP—Summary The minimum value of receiver window and congestion window is used as the size of the sliding window. If the network is in good condition, the TCP throughput should be in a steady level governed by the receiver window. TCP enters phase of Slow Start upon timeouts. If timeout frequently happens, TCP throughput will be dramatically cut down.

12. Sep 12, 2002Wireless Networking Seminar12 Simulation Environment Simulator: ns-2 with the wireless extension implemented by CMU. MAC Layer: IEEE 802.11 MAC Distributed Coordination function(DCF). Transport Layer: TCP connections carrying very large files. Network Environment –A Static String Network Topology –Interfering range is a little more than two times of the communication range 01234567 Interfering Range Communication Range

13. Sep 12, 2002Wireless Networking Seminar13 Instability Problem—Experiment Setup 12345 Source Destination A single TCP connection, with node 1 as the source and node 5 as the destination. Three sets of experiments with Maximum Window Size(window_) 32, 8, and 4 respectively.

14. Sep 12, 2002Wireless Networking Seminar14 Instability Problem—Experiment Result When window_=32 or 8, serious oscillation of throughput is observed. When window_4, throughput is stable.

15. Sep 12, 2002Wireless Networking Seminar15 Instability Problem—Trace Analysis(1) 12345 Data Ack RTS CTS Interfering Range of Node 2

16. Sep 12, 2002Wireless Networking Seminar16 Instability Problem—Summary Collision and exposed terminal problem prevent node 2 from receiving RTS from or sending CTS to node 1. The random back-off, big data packet, and sending back-to- back packets worsen the above problems. When window_ = 4, the chance to send back a CTS is greatly increased, so the throughput becomes stable. After node 1 fails seven times to receive CTS, node 1 believes there is a route failure and starts a route discovery. Before a route is available, node 1 can not send out a data packet. This period usually is long enough to cause a timeout at the TCP sender. For TCP, timeout triggers Slow Start, which significantly reduces the throughput.

17. Sep 12, 2002Wireless Networking Seminar17 Unfairness Problem—Experiment Setup 23456 Source Destination Source First SessionSecond Session In the first session, data flow from 6 to 4. In the second session, data flow from 2 to 3. The first session starts at 10.0s. The second session starts at 30.0s.

18. Sep 12, 2002Wireless Networking Seminar18 Unfairness Problem—Experiment Result(1) The first session has a throughput of about 450kbps from 10s to 30s, and 0kbps after 30s. The second session has a throughput of about 900kbps from 30s to 130s.

19. Sep 12, 2002Wireless Networking Seminar19 Unfairness Problem—Experiment Result(2) The first session never succeeds to send out packet with sequence number 2164.

20. Sep 12, 2002Wireless Networking Seminar20 Unfairness Problem—Trace Analysis(1) 23456 RTSData CTS Interfering Range of Node 5 Ack Interfering Range of Node 4 Data No Route

21. Sep 12, 2002Wireless Networking Seminar21 Unfairness Problem—Trace Analysis(2) 23456 RTSData CTS Interfering Range of Node 5 Ack Interfering Range of Node 4 Data No Route

22. Sep 12, 2002Wireless Networking Seminar22 Unfairness Problem—Summary In one-hop TCP connections, the interval between packet transmission is larger than that of the multi-hop TCP connections, which gives the one-hop connection more chances to transmit data. Random back-off is actually advantageous to the last succeeding host. The authors refer to this kind of unfairness problem “One-hop unfairness problem” and argue that since one-hop connection is common in a wireless network, one-hop unfairness problem can not be ignored.

23. Sep 12, 2002Wireless Networking Seminar23 Summary of The Paper Problems Shown: –Instability Problem –Unfairness problem Conclusions: –IEEE 802.11 does not work well in multi-hop wireless networks. –It may be inappropriate to take IEEE 802.11 as the MAC layer to simulate routing or transport protocols for multi- hop wireless networks.

24. Sep 12, 2002Wireless Networking Seminar24 Comments on The Paper Rooted in IEEE 802.11 MAC? –TCP is not designed with wireless networking in mind. –Timeout  Slow Start Instability problem can also be reduced to unfairness problem –They have almost the same network traces. –If we break down the network topology in the first problem, we obtain the topology in the second problem. Interfering range and communication range –If interfering range is the same as the communication range, the two problems presented in this paper will disappear. –Is the configuration of the interfering range simply an engineering issue?