1. The Impact of Multihop Wireless Channel on TCP Throughput and Loss Zhenghua Fu, Petros Zerfos, Haiyun Luo, Songwu Lu, Lixia Zhang, Mario Gerla INFOCOM2003, San Francisco, April 2003 Presented by Philip Hardebeck

2. Advanced Computer & Comm. Networks: Multihop Wireless Channel2 Outline n Introduction n Background n TCP Throughput –Several Topologies: Chain, Cross, Grid, Random n Simulations, Experiments, & Analysis n Proposed Solutions n Conclusions

3. Advanced Computer & Comm. Networks: Multihop Wireless Channel3 Introduction n Do TCP mechanisms work well for Wireless Multihop Networks (WMN)? n WMNs differ from wired networks. n There is an optimal TCP window size for a given topology and flow pattern.

4. Advanced Computer & Comm. Networks: Multihop Wireless Channel4 More Introduction n Packet losses increase as window size exceeds optimal, up to a threshold. n Link-RED and Adaptive Pacing are proposed to increase throughput.

5. Advanced Computer & Comm. Networks: Multihop Wireless Channel5 Background: MAC Basics ABCDE RTS CTS ABCDE DATA ACK ABCDE RTS ABCDE 8 x RTS CTS … random exponential backoff...

6. Advanced Computer & Comm. Networks: Multihop Wireless Channel6 Spatial Reuse and Contention ABCDEFGH Interfering RangeCommunication Range I ABCDE RTS CTS Interfering/Carrier Range of Node B RTS ABCDE DATA Interfering/Carrier Range of Node D

7. Advanced Computer & Comm. Networks: Multihop Wireless Channel7 TCP Throughput n Look at TCP throughput to show how well or poorly it performs spatial reuse. n Typical TCP operation doesn’t do a good job and the throughput is reduced. n Identify window size for highest throughput, and verify with hardware experiments.

8. Advanced Computer & Comm. Networks: Multihop Wireless Channel8 Chain Topology n Packets of a single flow interfere with one another. n Optimal window size is ~1/4 * number of hops in the chain.

9. Advanced Computer & Comm. Networks: Multihop Wireless Channel9 Optimal Window Size vs. Chain Length

10. Advanced Computer & Comm. Networks: Multihop Wireless Channel10 Throughput for 3 Packet Sizes

11. Advanced Computer & Comm. Networks: Multihop Wireless Channel11 Actual vs. Simulated Throughput

12. Advanced Computer & Comm. Networks: Multihop Wireless Channel12 Cross and Grid Topologies

13. Advanced Computer & Comm. Networks: Multihop Wireless Channel13 Aggregate Throughput and Window Size Table III

14. Advanced Computer & Comm. Networks: Multihop Wireless Channel14 Throughput Summary n Optimal window size exists for all topologies and flow patterns. n Optimal window size derivable only for simple configuration (chain). n Average TCP window size is much larger than optimal –Causes more packet drops and reduced throughput

15. Advanced Computer & Comm. Networks: Multihop Wireless Channel15 Loss Behavior n Buffer drop probability is not significant in WMN, but contention drop is. n “Network overload is no longer a bottleneck link property, but a shared feature of multiple links.” n Drop probability increases “gracefully” as load increases.

16. Advanced Computer & Comm. Networks: Multihop Wireless Channel16 TCP Packet Drop Probability

17. Advanced Computer & Comm. Networks: Multihop Wireless Channel17 UDP Packet Drop Probability at MAC layer

18. Advanced Computer & Comm. Networks: Multihop Wireless Channel18 Contrasting Drop Characteristics

19. Advanced Computer & Comm. Networks: Multihop Wireless Channel19 Analysis of Link Drop Probability n Modeling a random topology, drop probability is n Three regions of behavior –P l ~0: m, number of backlogged nodes, is < B*, maximum number of concurrent DATA transmitting nodes, and m~b~c

20. Advanced Computer & Comm. Networks: Multihop Wireless Channel20 Analysis of Link Drop Probability Continued n Other two regions: –P l increases linearly: m>B* and m<C*, maximum number of nodes with a clear channel –P l stable: m>C* - the amount of contention cannot increase

21. Advanced Computer & Comm. Networks: Multihop Wireless Channel21 Link-RED Algorithm

22. Advanced Computer & Comm. Networks: Multihop Wireless Channel22 Adaptive Pacing Algorithm

23. Advanced Computer & Comm. Networks: Multihop Wireless Channel23 TCP Throughput Comparison

24. Advanced Computer & Comm. Networks: Multihop Wireless Channel24 Multiflow TCP Throughput Comparison

25. Advanced Computer & Comm. Networks: Multihop Wireless Channel25 Average TCP Window Size Comparison

26. Advanced Computer & Comm. Networks: Multihop Wireless Channel26 Discussions n TCP Vegas doesn’t work as well as New Reno. n Optimal window sizes exist for flows with variable packet size, but more complicated. n LRED and Adaptive Pacing improve drop behavior.

27. Advanced Computer & Comm. Networks: Multihop Wireless Channel27 Related Work n Link-layer retransmission hides channel errors from upper layers n Dynamic ad hoc networks and link failure are studied (routing issues) n Studies of TCP ACK traffic using other MAC protocols n Capacity of ad hoc networks using UDP/CBR flows

28. Advanced Computer & Comm. Networks: Multihop Wireless Channel28 Conclusions n TCP throughput improves if the window size operates at optimal, maximizing channel spatial reuse. n TCP typically operates with a much larger window, reducing throughput. n Wireless nodes exhibit a graceful drop feature. n LRED and Adaptive Pacing improve throughput by up to 30%

29. Advanced Computer & Comm. Networks: Multihop Wireless Channel29 Problems/Weaknesses n No explanation for the 10% difference between simulation and experimental results. n Use of aggregate rate and window size makes it difficult to compare results to other papers’.

30. Advanced Computer & Comm. Networks: Multihop Wireless Channel30 Acknowledgements n Thanks to Professor Kinicki for the opportunity to make this presentation. n Thanks to Shugong Xu and Tarek Saadawi of CUNY for the MAC Basics and Spatial Reuse and Contention graphics.

31. Advanced Computer & Comm. Networks: Multihop Wireless Channel31 Questions/Comments?