TCP Services. Connection-oriented service Unicast communication only End-to-end guaranteed delivery Congestion control Flow control TCP 1

Transcription

1 TCP Services Connection-oriented service Unicast communication only End-to-end guaranteed delivery Congestion control Flow control TCP 1

2 TCP Handshake and Data Delivery TCP 2

3 Transport Layer E.g. HTTP (used by web services) typically uses TCP Client Server TCP Reliable transport between client and server required Steam oriented, not transaction oriented Network friendly: time-out congestion slow down transmission Well known TCP guesses quite often wrong in wireless and mobile networks Packet loss due to transmission errors TCP SYN TCP SYN/ACK TCP ACK HTTP request HTTP response GPRS: 500ms! Connection setup Data transmission >15 s no data Connection release Packet loss due to change of network Result Severe performance degradation Prof. Dr.-Ing. Jochen H. Schiller MC

4 Motivation I Transport protocols typically designed for Fixed end-systems Fixed, wired networks Research activities Performance Congestion control Efficient retransmissions TCP congestion control packet loss in fixed networks typically due to (temporary) overload situations router have to discard packets as soon as the buffers are full TCP recognizes congestion only indirect via missing acknowledgements, retransmissions unwise, they would only contribute to the congestion and make it even worse slow-start algorithm as reaction

5 Motivation II TCP slow-start algorithm sender calculates a congestion window for a receiver start with a congestion window size equal to one segment exponential increase of the congestion window up to the congestion threshold, then linear increase missing acknowledgement causes the reduction of the congestion threshold to one half of the current congestion window congestion window starts again with one segment TCP fast retransmit/fast recovery TCP sends an acknowledgement only after receiving a packet if a sender receives several acknowledgements for the same packet, this is due to a gap in received packets at the receiver however, the receiver got all packets up to the gap and is actually receiving packets therefore, packet loss is not due to congestion, continue with current congestion window (do not use slow-start)

6 TCP Congestion Control end-end control (no network assistance) transmission rate limited by congestion window size, Congwin, over segments: Congwin w segments, each with MSS bytes sent in one RTT: throughput = w * MSS RTT Bytes/sec TCP 6

7 AIMD TCP congestion avoidance: AIMD: additive increase, multiplicative decrease increase window by 1 per RTT decrease window by factor of 2 on loss event TCP Fairness Fairness goal: if N TCP sessions share same bottleneck link, each should get 1/N of link capacity TCP connection 1 TCP connection 2 bottleneck router capacity R TCP 7

8 TCP congestion control: probing for usable bandwidth: ideally: transmit as fast as possible (Congwin as large as possible) without loss increase Congwin until loss (congestion) loss: decrease Congwin, then begin probing (increasing) again two phases slow start congestion avoidance important variables: Congwin threshold: defines threshold between the slow start phase, congestion control phase TCP 8

9 RTT TCP Slowstart Slowstart algorithm Host A Host B initialize: Congwin = 1 for (each segment ACKed) Congwin++ until (loss event OR CongWin > threshold) exponential increase (per RTT) in window size (not so slow!) loss event: timeout (Tahoe TCP) and/or or three duplicate ACKs (Reno TCP) time TCP 9

10 TCP Congestion Avoidance Congestion avoidance /* slowstart is over */ /* Congwin > threshold */ Until (loss event) { every w segments ACKed: Congwin++ } threshold = Congwin/2 Congwin = 1 1 perform slowstart 1: TCP Reno skips slowstart (fast recovery) after three duplicate ACKs TCP 10

11 TCP State diagram (Reno) ACK Timeout/ w=1; ssth=w/2 ACK / w++ ACK Timeout/ w=1; ssth=w/2 Slow Start Congestion Avoidance W >= ssth ACK Timeout/ W=1; ssth=w/2 ACKn+1/ w=ssth Fast retransmit/ Fast Recovery w ACK / w++ 3 x ACKn ssth=w/2; w=ssth+3 1 TCP 11

12 TCP Congestion Control (New Reno) Slow-start Se w<ssth Ack: W++ T i m e o u t Slowstart Congest. avoidan. Se w<ssth: w Ack: W++ P. L o s s Fast retr/ Fast rec. Congesti avoidan. TCP 12

13 TCP Evolution Operating System TCP Implementation FreeBSD Reno FreeBSD 4.2 Reno FreeBSD 4.3 NewReno FreeBSD 4.4 NewReno FreeBSD 4.5 NewReno Windows 98 Tahoe sem Retransmissão Rápida Windows 2000 Tahoe sem Retransmissão Rápida RedHat 7.2 NewReno Tahoe Reno NewReno SACK (Selective Acknowledgement) FACK (Forward Acknowledgement) VEGAS TCP 13

14 Influences of mobility on TCPmechanisms TCP assumes congestion if packets are dropped May be wrong in wireless networks, if we often have packet loss due to transmission errors furthermore, mobility itself can cause packet loss, if e.g. a mobile node roams from one access point (e.g. foreign agent in Mobile IP) to another while there are still packets in transit to the wrong access point and forwarding is not possible Prof. Dr.-Ing. Jochen H. Schiller

15 TCP in UMTS TCP throughput as a function of the error rate and RLC layer retransmissions TCP 15

16 TCP in UMTS TCP delay (download) as a function of the error rate and number of retransmissions at the RLC layer TCP 16

17 TCP in UMTS TCP throughput as a function of the RTT TCP 17

18 TCP in UMTS TCP reception buffer occupancy As a function of the reception window size and the RTT TCP 18

19 Multipath TCP RFC 6824 (2013) Allows multimode terminals (e.g., smartphone) to take advantage of multiple interfaces at the same time, to increase the data rate. One MPTCP session bundles several TCP connections to transmit one stream. Sequence numbers are mapped between MPTCP stream and individual TCP streams. Supported in: Linux (and Android), Apple ios 7, etc.