1. ECEN4533 Data Communications Lecture #2711 March 2013 Dr. George Scheets n Problems Web 20-22 n Corrected Exams u One week after return (DL) n Design #2 u Due 15 March (Live) u Due 22 March (DL) n Quiz #2 on 27 March

2. ECEN4533 Data Communications Lecture #2813 March 2013 Dr. George Scheets n Read 8.5, 8.6 n Problems Web 23, Quiz #2 2010 n Design #2 u Due 15 March (Live) u Due 22 March (DL) n Quiz #2 on 27 March n Design 1 Final Scores u Hi = 70, Low = 38.5, Ave = 59.92, σ = 14.13

3. ECEN4533 Data Communications Lecture #2915 March 2013 Dr. George Scheets n Problems Quiz #2 2011 & 2011 n Design #2 u Due 15 March (Live) u Due 22 March (DL) n Quiz #2 on 27 March

4. Internet Packet Traffic (Daily) Amsterdam Internet Exchange

5. RFC2988 Algorithm time ACK # 3 5 RTO RTT Steady round-trip times of 500 msec with no variation. Initial RTO = 3 seconds. RTO settles on 500 msec.

6. RFC2988 Algorithm time ACK # 3 5 RTO RTT Round-trip times uniformly distributed between 0 & 1 second. Initial RTO = 3 seconds. RTO settles on about 1.3 seconds.

7. How does time out value effect performance the end user sees? n It doesn't if dropped packets are rare. u Line speed important, not RTT accuracy n Can severely effect wait if dropped packets are common.

8. TCP Flow Chart (simplified) slow start flow to send? normal ops congestion avoidance no yes

9. Terms n Advertised Window u What info sink is advertising in ACK n Available Window u Source's "Cleared to Send" moving window F Shrinks as segments sent F Expands as ACK's received n Congestion Window u Used during Slow Start & Congestion u Same as Advertised Window during Normal Ops

10. TCP Slow Start n Used on flow start up. n Congestion Window set = 1 segment n Info Source transmits 1 packet n Each ACK received bumps Congestion Window by +1 u Allows Source to transmit 2 packets for every ACK received (assuming Info Sink transmits an ACK for every packet it receives & nothing lost) n Congestion Window = Advertised Window? u Enter Normal Operation

11. TCP Slow Start ACK received? ConWind < AdvWind? normal ops yes no ConWind + 1 adjust window* adjust round-trip estimate & RTO congestion avoidance timed out? Units are segments. Congestion Window sets the pace. enter slow start ConWind = 1 window* = min(ConWind, AdvWind) send packet window*= window*-1 window* = 0? no yes halt window* = Available (moving) Window (at Source) continue slow start yes

12. TCP Normal Operation n Congestion Window = Advertised Window u During Normal Operation n Info Source u Transmits so long as Available Window > 0 n Info Sink u ACK's received segments u May adjust Window size n Transmission pace set by u Window Size as compared to NPD u Rate ACK's received

13. Normal Operation (Info Source) enter normal ops send packet decrease window* window* = 0? no yes ACK received? adjust window* yes window* = Available (moving) Window (at Source) halt congestion avoidance timed out? adjust round-trip estimate & RTO Normal Operations Congestion Window = Advertised Window

14. Time Out Occurs? Info Source... n Retransmits missing segment n Doubles RTO value u Readjusts RTO via RFC2988 when ACK received n Enters congestion avoidance mode u Sets Slow Start Threshold = Congestion Window/2 u Sets Congestion Window = 1 u Executes Slow Start F Congestion window +1 for every ACK received F Until Congestion Window > Slow Start Threshold, then... u Executes Additive Increase F Congestion Window +1 every RTT F Until Congestion Window = Advertised Window, then... u Enter Normal Operations

15. Congestion Avoidance ACK received? ConWind < SSThres? Additive Increase yes no ConWind + 1 adjust window* adjust round-trip estimate & RTO congestion avoidance timed out? Units are segments. Congestion Window sets the pace. enter slow start ConWind = 1 SSThreshold = AdvWind/2 window* = min(ConWind, AdvWind) send packet window*= window*-1 window* = 0? no yes halt window* = Available (moving) Window (at Source) continue slow start yes

16. Congestion Avoidance ACK received? ConWind < AdvWind? Normal Ops yes no adjust window* adjust round-trip estimate & RTO congestion avoidance timed out? Units are segments. Congestion Window sets the pace. Additive Increase window* = min(ConWind, AdvWind) send packet window*= window*-1 window* = 0? no yes halt window* = Available (moving) Window (at Source) continue Additive yes ConWind + 1 Every RTT

17. Fast Retransmission n 3 duplicate ACK's? u Immediately ship missing segment n No Time Out? u Proceed as if everything OK n Time out before missing segment ACKed? u Enter "Fast Recovery" u Congestion Window cut in half u Use Additive Increase Congestion Window +1 every RTT

18. Fast Recovery ACK received? ConWind < AdvWind? Normal Ops yes no adjust window* adjust round-trip estimate & RTO congestion avoidance timed out? Units are segments. Congestion Window sets the pace. Additive Increase window* = min(ConWind, AdvWind) send packet window*= window*-1 window* = 0? no yes halt window* = Available (moving) Window (at Source) continue Additive yes ConWind + 1 Every RTT

19. If line fails, fix at lowest layer n n SONET 1st u u Shoots for 50 msec fix n n ATM or Carrier Ethernet Second n n Routers 3rd

20. Two main types of protection n n Line Protection u u All traffic on inop line routed around break u u Treated as one bundle n n Path Protection u u All affected end-to-end connections are completely rerouted n n Line Protection is faster n n Path Protection uses less resources

21. Failure Analysis/Survivability

22. Anaheim to Dallas Link Fails 1-30 Paths 31-60 Paths 61-90 Paths 91-120 Paths 121+ Paths Figure 1) 17 Node, 27 Link Test Network with 500 End-to-End Optical Paths. Links are color coded based on traffic density. If Line Protection, all affected traffic is rerouted around break.

23. Anaheim - Dallas Fails Example here is Path Protection. Note changes in NE.

24. Path Reconfigure Time

25. UDP Header Source PortDestination Port Sequence Length Checksum