Ats Fit Gap Template - PowerPoint

Document Sample
Ats Fit Gap Template - PowerPoint Powered By Docstoc
					Measurement of One-Way Transit
Time in IP Routers

HET-NETs’05 Working Conference
18 – 20 July 2005
Ilkley, West Yorkshire, United Kingdom


Adrian Popescu and Doru Constantinescu
Dept. of Telecommunication Systems
Blekinge Institute of Technology
Karlskrona, Sweden
                                       Outline
                                        Introduction

                                        Router Architecture

                                        One-Way Transit Time

                                        Queueing Delay in Chained IP Routers

                                        Measurement Setup

                                        Estimation of OWTT and Other Router Delays
    Blekinge Institute of Technology
Dept. of Telecommunication Systems      Sources of Errors
              SE-371 79 Karlskrona
                 +46 455 38 50 00
              www.ats.tek.bth.se
                                        Modeling Methodology
                                       Outline (cont.)
                                        Experiments

                                        Processing Delay of a Router

                                        Router Delay for a Single Data Flow

                                        Router Delay for More Data Flows

                                        End-to-End Delay for a Chain of Routers

                                        Conclusions
    Blekinge Institute of Technology
Dept. of Telecommunication Systems      Future Work
              SE-371 79 Karlskrona
                 +46 455 38 50 00
              www.ats.tek.bth.se
                                       Introduction
                                        Measurements of One-Way Transit Time
                                         (OWTT) and other router delays

                                        Goals:
                                             Design of a measurement system to follow specifications of
                                              IETF RFC 2679

                                             Delay measurements

                                             Understanding the delay process in IP routers



    Blekinge Institute of Technology
Dept. of Telecommunication Systems
              SE-371 79 Karlskrona
                 +46 455 38 50 00
              www.ats.tek.bth.se
                                       Router Architecture




    Blekinge Institute of Technology
Dept. of Telecommunication Systems
              SE-371 79 Karlskrona      Basic activities
                 +46 455 38 50 00
              www.ats.tek.bth.se
                                            Routing
                                            Datagram forwarding
                                       One-Way Transit Time
                                          OWTT has several components
                                                                              N
                                                      OWTT  D prop   Dn ,i
                                                                             i 0

                                           where the delay per node i

                                                       Dn ,i  Dtr ,i  D proc,i  Dq ,i

                                          OWTT can be partitioned into a deterministic component and a
                                           stochastic component
    Blekinge Institute of Technology
Dept. of Telecommunication Systems
              SE-371 79 Karlskrona
                                                       OWTT  Dd  Ds
                 +46 455 38 50 00
              www.ats.tek.bth.se
                                       Queueing Delay in Chained IP Routers



                                          Fundamental problem
                                               Traffic merging

                                          Main consequences
                                               Character of arrival process at a downstream queue changes
                                               Appearance of correlations

                                          Important classes of correlations
                                               Autocorrelations in packet interarrival times
                                               Autocorrelations in packet service times
                                               Crosscorrelations between packet interarrival times and packet service times
                                               Crosscorrelations between packet service times in tandem queues
    Blekinge Institute of Technology
Dept. of Telecommunication Systems        Another important consequence
              SE-371 79 Karlskrona
                 +46 455 38 50 00
                                               Almost impossible to do precise queueing analysis
              www.ats.tek.bth.se
                                          Actual solution used
                                               Kleinrock independence assumption
                                       Measurement Setup




                                          Dedicated Measurement Points (MPs) equipped with (synchronized)
    Blekinge Institute of Technology
                                           DAG 3.5E
Dept. of Telecommunication Systems
              SE-371 79 Karlskrona        Control in generating and capturing network traffic
                 +46 455 38 50 00
              www.ats.tek.bth.se          UDP traffic generated with TCP-like characteristics
                                          High accuracy of timestamps
                                          Off-line data analysis
                                       Measurement Setup (cont.)
                                          Control of link utilization Lu and Hurst parameter H
                                               Pareto distributed traffic for the packet length generated at the application level
                                                with the shape parameter α
                                               Inter packet gap-time exponentially distributed with parameter λ
                                               Number of traffic sources n

                                          Traffic generation
                                               Traffic generators
                                               Generated traffic type
                                                     World Wide Web-like traffic at the application level
                                                     Fractional Brownian Motion (fBm) at the network level

                                          Packet identification
    Blekinge Institute of Technology
Dept. of Telecommunication Systems             Hashing and masking
              SE-371 79 Karlskrona
                 +46 455 38 50 00                    SHA-1 algorithm
              www.ats.tek.bth.se
                                               Packet matching
                                               Use of template containers defined by the Standard Template Library
                                       Estimation of OWTT and Other Router Delays




    Blekinge Institute of Technology
Dept. of Telecommunication Systems
              SE-371 79 Karlskrona
                 +46 455 38 50 00
              www.ats.tek.bth.se
                                       Estimation of OWTT and Other Router Delays
                                       (cont.)
                                          Matrices used in estimation of different delays
                                               Timestamps for packet n captured by DAGi [Ti(n)]

                                               Interarrival times for packet n captured by DAGi
                                                [IntArri(n)]=[Ti(n)-Ti(n-1)]

                                               Service times for packet n [Serv(n)]

                                               One-Way Transit Times for packet n measured between DAGj and
                                                DAGi [OWTTij,i(n)]=[Tj(n)-Ti(n)]

                                               Router transit times for packet n [RTT(n)]=[OWTTj,i(n)-Serv(n)]

                                               Minimum delay for a specific packet size L [Dmin|ln=L]

    Blekinge Institute of Technology           Queueing delay for packet n [Queue(n)]=[RTT(n)-Dmin|ln=L]
Dept. of Telecommunication Systems
              SE-371 79 Karlskrona
                 +46 455 38 50 00
              www.ats.tek.bth.se
                                       Sources of Errors
                                          Duplicate packets
                                              Very low probability of occurrence, due to strictly controlled environment as
                                               well as own generated traffic

                                          Unmatched packets
                                              Mostly because of other interfering traffic, e.g., ARP and inter-router traffic
                                               as well as because of congestion avoidance in the router during heavy-load
                                               traffic conditions

                                              Low probability of occurrence, 0.01% to 5% for more than one million
                                               packets processed




    Blekinge Institute of Technology
Dept. of Telecommunication Systems
              SE-371 79 Karlskrona
                 +46 455 38 50 00
              www.ats.tek.bth.se
                                       Modeling Methodology
                                        Selection of candidate distribution(s)
                                             Use of visual techniques (CCDF plots, EDF plots, PDF plots, Hill plots,
                                              a-estimation plots)

                                             Determining whether a single or mixture of distributions is required

                                        Parameter estimation
                                             Maximum Likehood Estimation (MLE) method

                                             Use of successive right censoring in the case of mixture of distributions

                                        Fitness assessment
                                             Goodness-of-fit significance tests (null hypothesis): Kolmogorov-Smirnov, l2
    Blekinge Institute of Technology          and Anderson-Darling can be used
Dept. of Telecommunication Systems
              SE-371 79 Karlskrona
                                             Drawback: they always tend to reject the null hypothesis in the case of large
                 +46 455 38 50 00
              www.ats.tek.bth.se
                                              sample

                                             Own developed method (David Erman): similar to the EDF test, but it does
                                              not suffer as much with increasing size of sample space
                                       Experiments
                                          Classes of OWTT experiments

                                               One router with single data flow

                                               One router with more data flows

                                               Chain of routers with more data flows

                                          9 experiments done for each class of experiments with
                                           different H, Lu and combinations of traffic mixture




    Blekinge Institute of Technology
Dept. of Telecommunication Systems
              SE-371 79 Karlskrona
                 +46 455 38 50 00
              www.ats.tek.bth.se
                                       Experiments (cont.)




    Blekinge Institute of Technology
Dept. of Telecommunication Systems
              SE-371 79 Karlskrona
                 +46 455 38 50 00
              www.ats.tek.bth.se
                                       Processing Delay of a Router




    Blekinge Institute of Technology     Example of CISCO 3620 router processing delay for
                                         ICMP and UDP payloads
Dept. of Telecommunication Systems
              SE-371 79 Karlskrona
                 +46 455 38 50 00
              www.ats.tek.bth.se
                                       Processing Delay of a Router (cont.)




                                         Minimum router transit time for a specific packet size
    Blekinge Institute of Technology     [Dmin|ln=L] obtained, with 95% confidence bounds, in
Dept. of Telecommunication Systems
              SE-371 79 Karlskrona
                                         experiments 1-3 and 1-7
                 +46 455 38 50 00
              www.ats.tek.bth.se
                                       Router Delay for a Single Data Flow




                                              Measurement configuration
    Blekinge Institute of Technology
Dept. of Telecommunication Systems
              SE-371 79 Karlskrona
                 +46 455 38 50 00
              www.ats.tek.bth.se
                                       Router Delay for a Single Data Flow (cont.)




    Blekinge Institute of Technology
Dept. of Telecommunication Systems
              SE-371 79 Karlskrona
                 +46 455 38 50 00
              www.ats.tek.bth.se
                                       Router Delay for a Single Data Flow (cont.)




                                          Main observations:
    Blekinge Institute of Technology
                                              Limited disparity in summary statistics
Dept. of Telecommunication Systems
              SE-371 79 Karlskrona            Variance and mean slightly increasing with Lu and H
                 +46 455 38 50 00
              www.ats.tek.bth.se              One delay has a maximum that is unusual large (55.5 ms)
                                       Router Delay for a Single Data Flow (cont.)




    Blekinge Institute of Technology
Dept. of Telecommunication Systems
              SE-371 79 Karlskrona
                 +46 455 38 50 00
              www.ats.tek.bth.se


                                           Delay distributions obtained in experiment 1-9
                                       Router Delay for a Single Data Flow (cont.)




                                          Modeling results obtained for delays in experiment 1-9
    Blekinge Institute of Technology
Dept. of Telecommunication Systems
              SE-371 79 Karlskrona
                 +46 455 38 50 00
              www.ats.tek.bth.se
                                       Router Delay for a Single Data Flow (cont.)




    Blekinge Institute of Technology
Dept. of Telecommunication Systems
              SE-371 79 Karlskrona
                 +46 455 38 50 00
              www.ats.tek.bth.se
                                           Modeling of delays obtained in experiment 1-9
                                       Router Delay for a Single Data Flow (cont.)




                                       RTT  p1( Lu, H ) min D  [1  p1( Lu, H )]Queueing owtt  p 2( Lu, H )serv  [1  p 2( Lu, H )]RTT
    Blekinge Institute of Technology
Dept. of Telecommunication Systems
              SE-371 79 Karlskrona
                 +46 455 38 50 00
              www.ats.tek.bth.se
                                       Router Delay for a Single Data Flow (cont.)




    Blekinge Institute of Technology
Dept. of Telecommunication Systems
                                          Results obtained in experiment 1 on power spectrum
              SE-371 79 Karlskrona
                 +46 455 38 50 00
              www.ats.tek.bth.se
                                       Router Delay for More Data Flows




    Blekinge Institute of Technology
Dept. of Telecommunication Systems
              SE-371 79 Karlskrona
                 +46 455 38 50 00             Measurement configuration
              www.ats.tek.bth.se
                                       Router Delay for More Data Flows (cont.)




    Blekinge Institute of Technology
Dept. of Telecommunication Systems
              SE-371 79 Karlskrona
                 +46 455 38 50 00
              www.ats.tek.bth.se
                                       Router Delay for More Data Flows (cont.)




                                           Main observations:
    Blekinge Institute of Technology
Dept. of Telecommunication Systems
              SE-371 79 Karlskrona
                                                Larger disparity for some OWTT statistics
                 +46 455 38 50 00
              www.ats.tek.bth.se                Samples with large delays are more common

                                                Heavy tail observed in histograms (dependent on Lu and H)
                                       Router Delay for More Data Flows (cont.)




                                              Modeling results obtained in experiment 2-5
    Blekinge Institute of Technology
Dept. of Telecommunication Systems
              SE-371 79 Karlskrona
                 +46 455 38 50 00
              www.ats.tek.bth.se
                                       Router Delay for More Data Flows (cont.)




    Blekinge Institute of Technology
Dept. of Telecommunication Systems
              SE-371 79 Karlskrona
                 +46 455 38 50 00
              www.ats.tek.bth.se

                                          Modeling of delays obtained in experiment 2-5
                                       Router Delay for More Data Flows (cont.)




                                       RTT  p1( Lu, H ) min D  [1  p1( Lu, H )]Queueing owtt  p 2( Lu, H )serv  [1  p 2( Lu, H )]RTT
    Blekinge Institute of Technology
Dept. of Telecommunication Systems
              SE-371 79 Karlskrona
                 +46 455 38 50 00
              www.ats.tek.bth.se
                                       End-to-End Delay for a Chain of Routers




    Blekinge Institute of Technology
Dept. of Telecommunication Systems
              SE-371 79 Karlskrona
                 +46 455 38 50 00
              www.ats.tek.bth.se
                                                Measurement configuration
                                       End-to-End Delay for a Chain of Routers (cont.)




    Blekinge Institute of Technology
Dept. of Telecommunication Systems
              SE-371 79 Karlskrona
                 +46 455 38 50 00         Inte mycket text här
              www.ats.tek.bth.se
                                       End-to-End Delay for a Chain of Routers (cont.)




                                           Main observations:
    Blekinge Institute of Technology
Dept. of Telecommunication Systems
              SE-371 79 Karlskrona
                                                Large disparity for all statistics (except minimum)
                 +46 455 38 50 00
              www.ats.tek.bth.se                Large number of samples with large delays

                                                Heavy tail observed in histograms (dependent on Lu and H)
                                       End-to-End Delay for a Chain of Routers (cont.)




    Blekinge Institute of Technology
Dept. of Telecommunication Systems
              SE-371 79 Karlskrona
                 +46 455 38 50 00
              www.ats.tek.bth.se




                                        Modeling results obtained in experiment 3-7 at routers R1, R2 and R3
                                       End-to-End Delay for a Chain of Routers (cont.)




    Blekinge Institute of Technology   OWTT distributions obtained in experiment 3-7 at routers R1, R2 and R3
Dept. of Telecommunication Systems
              SE-371 79 Karlskrona
                 +46 455 38 50 00
              www.ats.tek.bth.se
                                       End-to-End Delay for a Chain of Routers (cont.)




    Blekinge Institute of Technology
Dept. of Telecommunication Systems
              SE-371 79 Karlskrona
                 +46 455 38 50 00
              www.ats.tek.bth.se




                                             Summary of measured OWTT performance
                                       Conclusions
                                          Dedicated measurement system for delay measurements
                                           in IP routers (IETF RFC2679)

                                          Measurement study of delay through IP routers

                                          We confirm earlier results about the dependency of
                                           delay on traffic characteristics, link conditions, hardware
                                           implementations and IOS releases

                                          New results indicate that the delay in IP routers can be
                                           well modeled with the help of three distributions

    Blekinge Institute of Technology
Dept. of Telecommunication Systems
              SE-371 79 Karlskrona
                 +46 455 38 50 00
              www.ats.tek.bth.se
                                       Future Work
                                          Analytical models for OWTT, to include possible correlations
                                           between packet service times at adjacent nodes

                                          E2E delay formula in a chain of IP routers

                                          Targets

                                           RTT  p1( Lu, H ) min D  [1  p1( Lu, H )]Queueing

                                           owtt  p 2( Lu, H )serv  [1  p 2( Lu, H )]RTT

                                           owtt _ e 2 e  r1  r 2  r 3  ...... rn  ci (ri, rj )
    Blekinge Institute of Technology
Dept. of Telecommunication Systems
              SE-371 79 Karlskrona
                 +46 455 38 50 00
              www.ats.tek.bth.se

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:18
posted:12/21/2010
language:English
pages:39
Description: Ats Fit Gap Template document sample