1. Transport Protocols for Wireless Ad Hoc Networks 1

2. TCP Congestion Control EventStateTCP Sender ActionCommentary ACK receipt for previously unacked data SSCongWin = CongWin + MSS, If (CongWin > Threshold) set state to “Congestion Avoidance” Resulting in a doubling of CongWin every RTT ACK receipt for previously unacked data CACongWin = CongWin+MSS * (MSS/CongWin) Additive increase, resulting in increase of CongWin by 1 MSS every RTT Loss event detected by triple duplicate ACK SS or CAThreshold = CongWin/2, CongWin = Threshold, Set state to “Congestion Avoidance” Fast recovery, implementing multiplicative decrease. CongWin will not drop below 1 MSS. TimeoutSS or CAThreshold = CongWin/2, CongWin = 1 MSS, Set state to “Slow Start” Enter slow start Duplicate ACKSS or CAIncrement duplicate ACK count for segment being acked CongWin and Threshold not changed loss event = timeout or 3 duplicate acks 2

3. Problems in TCP over Wireless Ad Hoc Networks  Miss interpretation of packet loss  TCP interprets any packet loss as a sign of congestion. TCP sender reduces congestion window.  On wireless links, packet loss can also occur due to random channel errors, handoffs or route changes. Not due to congestion. Reducing window may be too conservative. Leads to poor throughput.  How to distinguish loss due to congestion from loss due to other wireless/mobility reasons?  Frequent path break leads to throughput degradation  Route reestablishment delay > RTO  congestion notification  reduce CW  performance degradation 3

4. Factors that affect TCP performance  Wireless transmission errors  Multi-hop routes on shared wireless medium  For instance, adjacent hops typically cannot transmit simultaneously  Route failures due to mobility  Unfairness 4

5. Wireless Transmission Errors  If # of errors is small  may be corrected by an error correcting code  delivered  Excessive bit errors  result in a packet being discarded, possibly before it reaches the transport layer  lost  Random errors may cause Fast Retransmission  retransmission of lost packet  reduction in congestion window  reduces the throughput (unnecessary)  Sometimes congestion response may be appropriate  When a channel is in a bad state for a long duration, it might be better to let TCP backoff  Burst errors may cause Timeouts  If wireless link remains unavailable for extended duration, a window worth of data may be lost driving through a tunnel passing a truck  Timeout results in slow start  Slow start reduces congestion window to 1 MSS, reducing throughput  Reduction in window in response to errors unnecessary 5

6. TCP on Multi-hop Wireless  TCP throughput drops with increase in #hops.  Reasons  Each hop adds additional self-contention. Also, more delay, more variability in delay, and more chance of packet loss. They increase RTO.  Packet transmission can occur on at most one hop among 3 consecutive hops  Contention between TCP Data and ACKs traveling in opposite directions 6

7. Impact of Mobility: TCP Throughput  Think of a TCP session where one end point is a mobile.  Route recomputations cause interruptions often longer than RTO.  Source doubles RTO.  Large original RTO or longer interruptions especially vulnerable.  Explicit notifications have been found to be useful.  Explicit route failure and route reconnect notifications to TCP source 7

8. Impact of MAC Unfairness ACK Traffic TAP1 TAP2 TAP3 TAP4  Notations: Obj = objective throughput for fair sharing Basic = no RTS/CTS  Two longer flows starve  Similar effect regardless whether RTS/CTS are used 8

9. How to Improve TCP Throughput  Mask wireless loss from the TCP sender.  TCP sender will not reduce congestion window  Explicitly notify the TCP sender about cause of packet loss.  TCP sender will not reduce congestion window for wireless losses.  Solutions may be at the sender, at the receiver, or at an intermediate node (basestation). 9

10. TCP-F: TCP Feedback [9-3]  Aims to minimize the throughput degradation resulting from the frequent path breaks  Allows the source to be informed of a route disconnection as a result of node mobility.  Avoid going through SS process  Failure point (FP) detecting link break  Originate Route Failure Notification(RFN) packet toward the source  Every intermediate node  updates its routing table and forwards toward the source,  Or, if has an alternate route to the destination, discards RFN and use the alternate route.  When the source receives RFN, it enters SNOOZ state  The source stops transmitting all data packets  It freeze all timers, the current cwnd size, and value of other state variables, such as RTT estimate, RTO.. It then initiates route failure timer: depend on the worst-case route repair time  When the source receives route reestablishment notification (RRN) packet, or route failure timer is expired,  Data transmission will be resumes and all timers and variables will be restored  Do FP or intermediate nodes always find the path to the source ? 10

11. TCP-ELFN: TCP with Explicit Link Failure Notification [9-8]  Similar to TCP-F, except for  Handling ELFN  Orignates “Destination Unreachable” ICMP error msg, or  Piggy back ELFN on RERR message that is sent to the sender  Detecting route reestablishment  When the TCP sender receives the ELFN, Disables RTO, and enters a standby state Periodically originates probe packets to see if a new route is reestablished If receives an ACK for the probe, leaves standby state, and restore RTO  Less dependent routing protocol 11

12. TCP-BuS: TCP with Buffering Capability and Sequence Information [9-10]  Use explicit feed back, but more dependent on routing protocol  ABR was proposed for underlying routing protocol  When route break is detected,  PN (pivot node): send Explicit Route Disconnection Notification(ERDN) to TCP-Bus sender buffers packets in transit from sender to PN until partial path is reestablished,  Downstream intermediate node: send Route Notification(RN) to TCP-BuS receiver Intermediate node that receives an RN packet discards all packets belonging to the flow  Routing protocol  LQ  destination: carries SEQ of TCP segment buffered at PN  REPLY  PN: carries SEQ of the last segment successfully received by receiver  When route is repaired,  PN: send Explicit Route Successful Notification(ERSN) to TCP-BuS  Sender: understands The last successfully received packet at destination The packets buffered at PN, the packets lost in transition  Receiver understands that lost packets will be delayed further To avoid unnecessary requests for fast retransmission, use selective ACK  Packets in transit before route disconnection ; Destination node continue to send ACK 12

13. ATCP: Ad Hoc TCP [9-12]  Compatible with TCP  Implemented only at TCP sender  New interpretation  ECN (explicit congestion notification) or ICMP source quench: congestion  ICMP DUR: link break  3 duplicated ACK: Avoid fast reTX and fast recovery (CW backoff) TCP state  persist state to avoid CW backoff reTx by ATCP 13

14. Split TCP [9-13]  To solve  the degradation of throughput with increasing path length  unfairness in wireless MAC (channel capture effect)  Lessen impact of mobility  Split a long TCP connection into a set of short concatenated TCP connections (called segments or zones)  To operate at its own transmission rate  Proxy node  Buffering  Local ACK  Split congestion control and end- to-end reliability  Congestion window: local ACK  End-to-end window: end-to-end ACK  Drawback: IP pay load encryption cannot be used 14