Data Communication And Computer Networks
Data Communication And Computer Networks CS 53600
Popular in Course
Popular in ComputerScienence
This 23 page Class Notes was uploaded by Nick Rowe on Saturday September 19, 2015. The Class Notes belongs to CS 53600 at Purdue University taught by Ramana Kompella in Fall. Since its upload, it has received 58 views. For similar materials see /class/208051/cs-53600-purdue-university in ComputerScienence at Purdue University.
Reviews for Data Communication And Computer Networks
Report this Material
What is Karma?
Karma is the currency of StudySoup.
You can buy or earn more Karma at anytime and redeem it for class notes, study guides, flashcards, and more!
Date Created: 09/19/15
CS 536 Data Communications and Computer Networks Lecture 12 TCP Fairness 10212008 CS 536 Computer Networks 1 Announcements CI Midterm Should we go for bestof two midterms 7 Cl Project Should we ignore project and adjust the weightoge to PA 2 7 CI Currently 0 Programming 30 10 20 0 Written Homeworks 20 o Midterm 20 0 Term Project 20 will be removed 0 Final 30 CS 536 Computer Networks 2 Midterm Ave 635 Med673 Max89 mum laguwa m 257a 156a 5a ma 55w Score Range awaza l ram 712D 5 2n 5 3n 5 4U 45 5D 55 an 55 7D 75 an 55 an an 75 1 Frequency CS 536 Compuvzr Networks 3 Midterm Problem Cl Assume ThaT There is only one rouTerswiTch beTween a source and desTinaTion The bandwidTh of The firsT link ie from source To The rouTer is R1 The second link ie from rouTer To desTinaTion is R2 WhaT is The ToTal Time iT Takes To Transfer K byTes from The source To The desTinaTion Ignore queuing propagaTion and processing delays 0 Here we need The boTTleneck bandwidTh Bmin R1R2 The endToend ThroughpuT is basically The boTTlenech bandwidTh Thus The Time iT Takes To Transfer K byTes is KB CS 536 CompuTer NeTworks 4 Recap TCP congestion control additive increase multiplicative decrease Cl Approachincrease Transmission ra re window size probing for usable bandwid rh un ril loss occurs 0 addifive increase increase CongWin by 1 M55 every RTT un ril loss de rec red o muHpicafive decrease cu r CongWin in half af rer loss 24Kbytes Saw TooTh behavior probing for bandwidTh 16 Kbytes 8 Kbytes congestion window size time This AIMD is also called TCP CongesTIon AVOIdance CS 536 Compufer Nefworks 5 Recap Summary TCP Congestion Control Cl When CongWin is below Threshold sender in slowstart phase window grows exponentially Cl When CongWin is above Threshold sender is in congestionavoidance phase window grows linearly Cl When a triple duplicate ACK occurs Threshold set to CongWin 2 and CongWin set to Threshold Cl When timeout occurs Threshold set to CongWin2 and CongWin is set to 1 M55 CS 536 Computer Networks 6 TCP sender congestion control State Event TCP Sender Action Commentary Slow Start ACK receipt CongWin CongWin MSS Resulting in a doubling of SS for previously If CongWin gt Threshold CongWin every RTT unacked set state to Congestion data Avoidance Congestion ACK receipt CongWin CongWinMSS Additive increase resulting Avoidance for previously MSSCongWin in increase of CongWin by CA unacked 1 MSS every RTT data SS or CA Loss event Threshold CongWin2 Fast recovery detected by CongWin Threshold implementing multiplicative triple Set state to Congestion decrease CongWin will not duplicate Avoidance drop below 1 MSS ACK SS or CA Timeout Threshold CongWin2 Enter slow start CongWin 1 MSS Set state to Slow Start SS or CA Duplicate Increment duplicate ACK count CongWin and Threshold not ACK for segment being acked changed CS 536 Computer Networks TCP Fairness Fairness goal if K TCP sessions share some bottleneck link of bandwidth R each should have average rate of R K TCP connection 1 bottleneck router capacity R TCP connection 2 CS 536 Computer Networks 8 Why is TCP fair Two competing sessions El Additive increase gives slope of 1 as throughout increases El multiplicative decrease decreases throughput proportionally equal bandwidth share loss decreaSe window by factor of 2 congestion avoidance additive increaSe loss decreaSe window by factor of 2 congestion avoidance additive increaSe Connection 2 throughput 7U Connection 1 throughput R CS 536 Computer Networks 9 Fairness more Fairness and UDP Fairness and parallel TCP CI Multimedia apps often m do not use TCP Cl nothing prevents app from 0 do not want rate openmg Parallel throttled by congestion connections between 2 control hOS l39S Cl Instead use UDP CI Web browsers do this 0 PUmP audioVideola F CI Example link of rate R Eggignlgggle39 To 6mm supporting 9 connections D Research area TCP o pipeagplgsks for 1 TCP gets fmendly 0 new app asks for 11 TCPs gets RZ CS 536 Computer Networks 10 Chapter 3 Summary Cl principles behind TransporT layer services 0 mulTiplexing demulTiplexing O reliable daTa Transfer 0 flow conTrol Next 0 C 9 5ll COMV39Ol Cl leaving The neTwork CI insTanTiaTion and edge applicaTion implemenTaTion in The TransporT layers Imer39nel Cl inTo The neTwork O UDP core 0 TCP CS 536 CompuTer NeTworks 11 Additional Material Refer to Peterson Davie Chapter 6 Congestion control Ed2 CS 536 Computer Networks 12 Congestion Avoidance Solutions Cl TCP reac rs ro conges rion affer i r rakes place The da ra ra re changes rapidly and The sys rem is barely s rable or is even uns rable Cl Can we predicf when conges rion is abou r To happen and avoid i r Eg by de rec ring rhe knee of The cu rve Average Packe l delay Load CS 536 Compu rer Ne rworks 13 Congestion Avoidance Schemes Cl Rou rerbased Conges rion Avoidance O DECbi r RouTers explicile noTify sources abouT congesTion 0 Random Early De rec rion RED RouTers implicile noTify sources by dropping packeTs RED drops packe rs a r random and as a func rion of The level of congesTion CI Hos rbased Conges rion Avoidance O Source moni rors changes in RTT ro de rec r onse r of conges rion o A varie ry of algori rhms are described in Sec rion 643 of Pe rerson Davie ed 2 CS 536 Compu rer Ne rworks 14 DECbit CI Each packeT has a CongesTion NoTificaTionquot biT called The DECbiT in iTs header CI If any rouTer on The paTh is congesTed iT seTs The DECbiT o SeT if average queue lengTh gt 1 packeT averaged since The sTarT of The previous busy cycle Cl To noTify The source The desTinaTion copies DECbiT inTo ACK packeTs Cl Source adjusTs raTe To avoid congesTion o CounTs fracTion of DECbiTs seT in each window 0 If 50 seT increase raTe addiTiver o If gt50 o seT decrease raTe mulTipIicaTiver A Queue Length at router gt Time Averaging period gt DUDE Random Early Detection RED RED is based on DECbiT and was designed To work well wiTh TCP RED implicile no rifies sender by dropping packe rs Drop probabiliTy is increased as The average queue lengTh increases ExponenTial moving average of The queue lengTh is used so as To deTecT long Term congesTion ye r allow shorT Term bursTs To arrive Angenn1 1 a x Angenn a x Length i e Angenn1 2 Lengtha1 a CS 536 Compu rer Ne rworks 16 Random Early Detection RED packet no AVngize gt Mint Admit the yes new packet AVgQSize gt Maxth a no v v Admit packet with Drop the new packet a probability p RED Drop Probabilities Dz A I maxP If minT h lt Angen lt maxT h V 39 Av Len pAv Len maxP Angen mmTh minTh maxTh g g maxT h minT h quot count coun rs how Ion we39ve been in minTh lt Angenlt maxTh PrDr op Packe r w 9 1 countx Angen since we las r dropped a packet ie drops are spaced out in Time reducing likelihood of r39een39rer39ing slowstart CS 536 Compu rer Networks 18 Properties of RED CI Drops packe rs before queue is full in The hope of reducing rhe ra res of some flows CI Drops packe r for each flow roughly in propor rion ro i rs ra re CI Drops are spaced our in Time CI Because if uses average queue leng rh RED is Toleran r of burs rs Cl Random drops hopefully desynchronize TCP sources CS 536 Compu rer Ne rworks 19 Synchronization of sources 39 Source A AT TIHT TH CS 536 Compu rer Ne rwor39ks 20 Synchronization of sources Aggrega re Flow RTT Avg l l l l CS 536 Compu rer Ne rwor39ks 21 Desynchronized sources CS 536 Compu rer Ne rwor39ks 22 Desynchronized sources Aggrega re Flow NxRTT AVE a 2 g I CS 536 Compu rer Ne rwor39ks 23
Are you sure you want to buy this material for
You're already Subscribed!
Looks like you've already subscribed to StudySoup, you won't need to purchase another subscription to get this material. To access this material simply click 'View Full Document'