TCP Round-trip Times (RTTs RTTs) ) TCP Round-trip Times ( Popular - PowerPoint PPT Presentation

The UNIVERSITY The UNIVERSITY of of NORTH CAROLINA NORTH CAROLINA at at CHAPEL HILL CHAPEL HILL TCP Round-trip Times (RTTs RTTs) ) TCP Round-trip Times ( Popular belief: Popular belief: RTTs do not vary significantly within TCP connections RTTs do not vary significantly within TCP connections Variability in TCP Round-trip Times Variability in TCP Round-trip Times − Mean RTT can be used to approximate per-segment − Mean RTT can be used to approximate per-segment RTTs RTTs � TCP throughput models TCP throughput models Jay Aikat Aikat Jasleen Kaur Jasleen Kaur Don Smith Don Smith Kevin Kevin Jeffay Jeffay � Jay [Altman00, Kumar98, Lakshman97, Mathis97, Padhye98, … …] ] [Altman00, Kumar98, Lakshman97, Mathis97, Padhye98, Department of Computer Science Department of Computer Science � Analysis based on TCP behavior Analysis based on TCP behavior � University of North Carolina at Chapel Hill University of North Carolina at Chapel Hill [Zhang02] [Zhang02] http://www.cs http://www. cs. .unc unc. .edu edu/~ /~jasleen jasleen/research/RTT /research/RTT − RTT of initial segments is − RTT of initial segments is “ “typical typical” ” for a TCP connection for a TCP connection [Jiang02] [Jiang02] Do per-segment RTTs RTTs vary significantly within a TCP connection? vary significantly within a TCP connection? Do per-segment Do per-segment RTTs vary significantly within a TCP connection? 1 1 2 2 Data Source Data Source Data Source Data Source Regional ibiblio web Regional ibiblio web health-care servers health-care servers center center DAT DAT UNC Campus Department s UNC Campus Rest of Rest of UNC Campus Department s UNC Campus UNC UNC the world Research the world ISP ISP ACK Research ACK Residence Residence labs router labs router halls halls Wireless Monitor Monitor Wireless VPNs users VPNs users Modem Modem Rich data source! Rich data source! Rich data source! Rich data source! Rich data source! Rich data source! 3 3 4 4

Extracting Valid RTT Samples Extracting Valid RTT Samples Trace Statistics Trace Statistics Remote end-point Monitor d a t [ i ] Remote Remote Connections Connections RTT samples RTT samples Bytes Bytes Packets Packets ack [i] hosts hosts � Guiding principle: Guiding principle: � − − Consider only those Consider only those RTTs RTTs where there is where there is All All 22.7 million 22.7 million 962 K 962 K 252 million 252 million 628 GB 628 GB 511 M 511 M d a t unambiguous correspondence between an unambiguous correspondence between an connections connections [ i + 1 ] ACK and the DAT that triggered it. ACK and the DAT that triggered it. Connections Connections d a t [ i + 2 ] with at least with at least 1.1 million 1.1 million 258 K 258 K 236 million 236 million 581 GB 581 GB 464 M 464 M ? 10 samples 10 samples ? � Caveat: delayed Caveat: delayed ACKs ACKs � − − Could add 200 Could add 200 – – 500 ms to RTT estimates 500 ms to RTT estimates d a t [ i + 1 ] Large data set! Large data set! Large data set! ack [i+2] 5 5 6 6 Variability Across Connections Variability Across Connections Variability Within Connections Variability Within Connections � 60% connections see min RTT less than 100 ms 60% connections see min RTT less than 100 ms � � Median RTT: Median RTT: � − Only 23% see max RTT less than 100 ms − Only 23% see max RTT less than 100 ms − 30% of connections see a median RTT more than twice the min RTT − 30% of connections see a median RTT more than twice the min RTT � ACKs ACKs can arrive more than 25 s after DAT transmission! can arrive more than 25 s after DAT transmission! � � 90% RTT: 90% RTT: � − − 22% of connections see a 90% RTT more than 5 times the min RTT 22% of connections see a 90% RTT more than 5 times the min RTT � Mean and median Mean and median RTTs RTTs are comparable measures are comparable measures � − − 90% RTT increases with min RTT 90% RTT increases with min RTT 7 7 8 8

/ (SYN+ACK) RTT The SYN / (SYN+ACK) RTT The SYN Per-Segment RTTs Per-Segment RTTs: Mean or Distributions? : Mean or Distributions? � Is mean RTT a good approximation for per-segment Is mean RTT a good approximation for per-segment RTTs RTTs? ? � − TCP analytical models − TCP analytical models − TCP evaluation (simulations) − TCP evaluation (simulations) � RTT yielded by the SYN and SYN+ACK pair RTT yielded by the SYN and SYN+ACK pair � − Differs by more than 10% from min RTT for 14% of connections − Differs by more than 10% from min RTT for 14% of connections − Differs by more than 10% from median RTT for 50% of connections − Differs by more than 10% from median RTT for 50% of connections 9 9 10 10 Ongoing Work Ongoing Work Variability within connections Variability within connections � Impact of RTT variability on past work Impact of RTT variability on past work � − TCP analytical models TCP analytical models − − − Delay-based congestion control Delay-based congestion control − TCP evaluation (simulations) TCP evaluation (simulations) − − − TCP-based analysis TCP-based analysis � Causes of variability Causes of variability � − Congestion? Congestion? − Standard deviation in per-connection RTTs Standard deviation in per-connection RTTs Inter-quartile range Inter-quartile range − Increases rapidly in the range: − Increases rapidly in the range: − Increases consistently with − Increases consistently with − End-hosts? − End-hosts? (med-min) RTT (med-min) RTT min RTT = 100 ms min RTT = 100 ms – – 1 s 1 s − Increases less rapidly in other regions − Increases less rapidly in other regions � Models for per-connection Models for per-connection RTTs RTTs � − Accurate simulation environments − Accurate simulation environments 11 11 12 12

Per-Segment RTTs Per-Segment RTTs: Mean or Distributions? : Mean or Distributions? � Is mean RTT a good approximation for Is mean RTT a good approximation for � per-segment RTT? per-segment RTT? − TCP analytical models − TCP analytical models − TCP evaluation (simulations) TCP evaluation (simulations) − 13 13