《计算机网络——自顶向下方法与Internet特色》幻灯片Lecture35.ppt

《《计算机网络——自顶向下方法与Internet特色》幻灯片Lecture35.ppt》由会员分享，可在线阅读，更多相关《《计算机网络——自顶向下方法与Internet特色》幻灯片Lecture35.ppt（26页珍藏版）》请在淘文阁 - 分享文档赚钱的网站上搜索。

1、Chapter 3 outline3.1 Transport-layer services3.2 Multiplexing and demultiplexing3.3 Connectionless transport:UDP3.4 Principles of reliable data transfer3.5 Connection-oriented transport:TCPsegment structurereliable data transferflow controlconnection management3.6 Principles of congestion control3.7

2、 TCP congestion control3.7 TCP congestion controlnTCP uses end-to-end congestion control rather then network-assisted congestion control,since the IP layer provides no explicit feedback to the end systems regarding network congestion.nThe approach taken by TCP is to have each sender limit the rate a

3、t which it sends traffic into its connection as a function of perceived network congestion.nQuestion:How limit the rate?congestion window(CongWin)How perceive a congestion?Timeout or 3 duplicate ACKsHow change the send rate?TCP Reno congestion control algorithm:TCP Congestion Control:detailssender l

4、imits transmission:LastByteSent-LastByteAcked minCongWin,RcvWindowRoughly,CongWin is dynamic,function of perceived network congestion(新增加的一个变量，叫做拥塞窗口;若RcvWindow足够大，仅受拥塞窗口控制)How does sender perceive congestion?loss event=timeout or 3 duplicate acksTCP sender reduces rate(CongWin)after loss eventthree

5、 mechanisms of TCP congestion control:AIMD slow startconservative after timeout eventsrate=CongWin RTT Bytes/sec1.TCP congestion control:additive increase,multiplicative decrease(AIMD)Approach:increase transmission rate(window size),probing for usable bandwidth,until loss occursadditive increase:inc

6、rease CongWin by 1 MSS every RTT until loss detectedmultiplicative decrease:cut CongWin in half after loss timecongestion window sizeSaw toothbehavior:probingfor bandwidth2.TCP Slow StartWhen connection begins,CongWin=1 MSSExample:MSS=500 bytes&RTT=200 msecinitial rate=20 kbpsavailable bandwidth may

7、 be MSS/RTTdesirable to quickly ramp up to respectable rateWhen connection begins,increase rate exponentially fast until first loss event连接的开始阶段采用指数的增长方式，即一个RTT后若没有丢失，则CongWin增加一倍，直到发生第一次丢失为止，降低一半。叫慢启动。TCP Slow Start(more)When connection begins,increase rate exponentially until first loss event:doub

8、le CongWin every RTTdone by incrementing CongWin for every ACK receivedSummary:initial rate is slow but ramps up exponentially fastHost Aone segmentRTTHost Btimetwo segmentsfour segments3.RefinementQ:When should the exponential increase switch to linear?A:When CongWin gets to 1/2 of its value before

9、 timeout.Implementation:Variable Threshold 来管理复杂的动态过程，开始时设置一个很大的值，65KB.At loss event,Threshold is set to 1/2 of CongWin just before loss event3.Refinement:inferring lossAfter 3 dup ACKs:CongWin is cut in halfwindow then grows linearlyBut after timeout event:CongWin instead set to 1 MSS;window then g

10、rows exponentiallyto a threshold,then grows linearlyq 3 dup ACKs indicates network capable of delivering some segmentsq timeout indicates a“more alarming”congestion scenarioPhilosophy:Summary:TCP Congestion ControlWhen CongWin is below Threshold,sender in slow-start phase,window grows exponentially.

11、When CongWin is above Threshold,sender is in congestion-avoidance phase,window grows linearly.When a triple duplicate ACK occurs,Threshold set to CongWin/2 and CongWin set to Threshold.When timeout occurs,Threshold set to CongWin/2 and CongWin is set to 1 MSS.TCP sender congestion controlStateEvent

12、TCP Sender Action CommentarySlow Start(SS)ACK receipt for previously unacked data CongWin=CongWin+MSS,If(CongWin Threshold)set state to“Congestion Avoidance”Resulting in a doubling of CongWin every RTTCongestionAvoidance(CA)ACK receipt for previously unacked dataCongWin=CongWin+MSS*(MSS/CongWin)Addi

13、tive increase,resulting in increase of CongWin by 1 MSS every RTTSS or CALoss event detected by triple duplicate ACKThreshold=CongWin/2,CongWin=Threshold,Set state to“Congestion Avoidance”Fast recovery,implementing multiplicative decrease.CongWin will not drop below 1 MSS.SS or CATimeoutThreshold=Co

14、ngWin/2,CongWin=1 MSS,Set state to“Slow Start”Enter slow startSS or CADuplicate ACKIncrement duplicate ACK count for segment being ackedCongWin and Threshold not changed4.TCP throughputWhats the average throughout of TCP as a function of window size and RTT?Ignore slow startLet W be the window size

15、when loss occurs.When window is W,throughput is W/RTTJust after loss,window drops to W/2,throughput to W/2RTT.Average throughout:.75 W/RTT5.TCP FuturesExample:1500 byte segments,100ms RTT,want 10 Gbps throughputRequires window size W=83,333 in-flight segmentsThroughput in terms of loss rate:L=210-10

16、 for 10GbpsNew versions of TCP for high-speed needed!Fairness goal:if K TCP sessions share same bottleneck link of bandwidth R,each should have average rate of R/KTCP connection 1bottleneckroutercapacity RTCP connection 23.7.1 TCP FairnessWhy is TCP fair?Two competing sessions:Additive increase give

17、s slope of 1,as throughout increasesmultiplicative decrease decreases throughput proportionally RRequal bandwidth shareConnection 1 throughputConnection 2 throughputcongestion avoidance:additive increaseloss:decrease window by factor of 2congestion avoidance:additive increaseloss:decrease window by

18、factor of 2Fairness(more)Fairness and UDP（UDP是不公平的，它没有限制地向网络发送流量）Multimedia apps often do not use TCPdo not want rate throttled by congestion controlInstead use UDP:pump audio/video at constant rate,tolerate packet lossResearch area:TCP friendly(限制UDP流量)Fairness and parallel TCP connections（这种方式是不公平

19、的）nothing prevents app from opening parallel connections between 2 hosts.Web browsers do this Example:link of rate R supporting 9 cnctions;new app asks for 1 TCP,gets rate R/10new app asks for 11 TCPs,gets R/2!3.7.2 Delay modelingQ:How long does it take to receive an object from a Web server after s

20、ending a request?Ignoring congestion,delay is influenced by:TCP connection establishmentdata transmission delayslow startNotation,assumptions:Assume one link between client and server of rate RS:MSS(bits)O:object size(bits)no retransmissions(no loss,no corruption)Window size:First assume:fixed conge

21、stion window,W segments Then dynamic window,modeling slow start1.Fixed congestion window(1)First case:完成第一窗口内的所有报文段发送前，收到了第一个报文段的应答WS/R RTT+S/R:ACK for first segment in window returns before windows worth of data sentdelay=2RTT+O/R假设窗口长度为42.Fixed congestion window(2)Second case:收到窗口中第一个报文段之前已经发完了第一个

22、窗口中的所有内容WS/R RTT+S/R:wait for ACK after sending windows worth of data sentdelay=2RTT+O/R+(K-1)S/R+RTT-WS/R假设窗口长度为2K是涉及一个对象的数据窗口的数量3.TCP Delay Modeling:Slow Start(1)Now suppose window grows according to slow start Will show that the delay for one object is:where P is the number of times TCP idles at

23、server:-where Q is the number of times the server idles if the object were of infinite size.-and K is the number of windows that cover the object.4.TCP Delay Modeling:Slow Start(2)Example:O/S =15 segments K=4 windows Q=2 P=minK-1,Q=2Server idles P=2 timesDelay components:2 RTT for connection estab a

24、nd request O/R to transmit object time server idles due to slow startServer idles:P=minK-1,Q times5.TCP Delay Modeling(3)6.TCP Delay Modeling(4)Calculation of Q,number of idles for infinite-size object,is similar.Recall K=number of windows that cover objectHow do we calculate K?7.HTTP ModelingAssume

25、 Web page consists of:1 base HTML page(of size O bits)M images(each of size O bits)Non-persistent HTTP:M+1 TCP connections in seriesResponse time=(M+1)O/R+(M+1)2RTT+sum of idle timesPersistent HTTP:2 RTT to request and receive base HTML file1 RTT to request and receive M imagesResponse time=(M+1)O/R

26、+3RTT+sum of idle timesNon-persistent HTTP with X parallel connectionsSuppose M/X integer.1 TCP connection for base fileM/X sets of parallel connections for images.Response time=(M+1)O/R+(M/X+1)2RTT+sum of idle timesHTTP Response time(in seconds)RTT=100 msec,O=5 Kbytes,M=10 and X=5For low bandwidth,

27、connection&response time dominated by transmission time.Persistent connections only give minor improvement over parallel connections.HTTP Response time(in seconds)RTT=1 sec,O=5 Kbytes,M=10 and X=5For larger RTT,response time dominated by TCP establishment&slow start delays.Persistent connections now

28、 give important improvement:particularly in high delaybandwidth networks.Chapter 3:Summaryprinciples behind transport layer services:multiplexing,demultiplexingreliable data transferflow controlcongestion controlinstantiation and implementation in the InternetUDPTCPNext:leaving the network“edge”(application,transport layers)into the network“core”