Panel Topic: After Long Range Dependency (LRD) discoveries, what are the lessons learned so far to provide QoS for Internet advanced applications David.

Slides:



Advertisements
Similar presentations
Doc.: IEEE /1216r1 Submission November 2009 BroadcomSlide 1 Internet Traffic Modeling Date: Authors: NameAffiliationsAddressPhone .
Advertisements

Traffic and routing. Network Queueing Model Packets are buffered in egress queues waiting for serialization on line Link capacity is C bps Average packet.
Network Design and Optimization Queueing Theory Overview Dr. Greg Bernstein Grotto Networking
Modeling, Simulation and Measurements of Queuing Delay under Long-tail Internet Traffic Michele Garetto Politecnico di Torino, Italy Don Towsley University.
Network and Service Assurance Laboratory Analysis of self-similar Traffic Using Multiplexer & Demultiplexer Loaded with Heterogeneous ON/OFF Sources Huai.
2  Something “feels the same” regardless of scale 4 What is that???
A New Approach for Accurate Modelling of Medium Access Control (MAC) Protocols Presenter: Moshe Zukerman ARC Centre for Ultra Broadband Information Networks.
1 Self-Similar Ethernet LAN Traffic Carey Williamson University of Calgary.
On the Self-Similar Nature of Ethernet Traffic - Leland, et. Al Presented by Sumitra Ganesh.
1 Giuliano Casale, Eddy Z. Zhang, Evgenia Smirni {casale, eddy, Speaker: Giuliano Casale Numerical Methods for Structured Markov Chains,
A Nonstationary Poisson View of Internet Traffic T. Karagiannis, M. Molle, M. Faloutsos University of California, Riverside A. Broido University of California,
Self-Similarity in Network Traffic Kevin Henkener 5/29/2002.
1 TCOM 501: Networking Theory & Fundamentals Lecture 7 February 25, 2003 Prof. Yannis A. Korilis.
THE TITLE OF YOUR PAPER Your Name Communication Networks Laboratory School of Engineering Science Simon Fraser University.
Estimating Congestion in TCP Traffic Stephan Bohacek and Boris Rozovskii University of Southern California Objective: Develop stochastic model of TCP Necessary.
CS 5253 Workshop 1 MAC Protocol and Traffic Model.
Buffer Management for Shared- Memory ATM Switches Written By: Mutlu Apraci John A.Copelan Georgia Institute of Technology Presented By: Yan Huang.
Network Planning & Capacity Management Frank Yeong-Sung Lin ( 林永松 ) Department of Information Management National Taiwan University Taipei, Taiwan, R.O.C.
Internet Traffic Modeling Poisson Model vs. Self-Similar Model By Srividhya Chandrasekaran Dept of CS University of Houston.
Self-Similarity of Network Traffic Presented by Wei Lu Supervised by Niclas Meier 05/
1 Chapters 9 Self-SimilarTraffic. Chapter 9 – Self-Similar Traffic 2 Introduction- Motivation Validity of the queuing models we have studied depends on.
SELF-SIMILAR INTERNET TRAFFIC AND IMPLICATIONS FOR WIRELESS NETWORK PERFORMANCE IN SUDAN Presented By HUDA M. A. EL HAG University Of Khartoum – Faculty.
References for M/G/1 Input Process
Traffic Modeling.
Traffic Modelling and Related Queueing Problems Presenter: Moshe Zukerman ARC Centre for Ultra Broadband Information Networks EEE Dept., The University.
1 Seminar Introduction to Traffic Engineering October 2009 Ernst Nordström
Doc.: IEEE /349r1 Submission November 2000 S. Deshpande, S. Kandala, SharpSlide 1 Models for MPEG2 and Video Conferencing Sachin Deshpande Srinivas.
Yoni Nazarathy Gideon Weiss University of Haifa Yoni Nazarathy Gideon Weiss University of Haifa On the Variance of Queueing Output Processes Haifa Statistics.
Network Analysis A brief introduction on queues, delays, and tokens Lin Gu, Computer Networking: A Top Down Approach 6 th edition. Jim Kurose.
Introduction to Queuing Theory
DELAYED CHAINING: A PRACTICAL P2P SOLUTION FOR VIDEO-ON-DEMAND Speaker : 童耀民 MA1G Authors: Paris, J.-F.Paris, J.-F. ; Amer, A. Computer.
DISPERSITY ROUTING: PAST and PRESENT Seungmin Kang.
Quality of Service in the Internet The slides of part 1-3 are adapted from the slides of chapter 7 published at the companion website of the book: Computer.
AIMS’99 Workshop Heidelberg, May 1999 Linking User Acceptance and Network Performance Miles Wilkins (BT) P807 (JUPITER2)
Yoni Nazarathy Gideon Weiss University of Haifa Yoni Nazarathy Gideon Weiss University of Haifa The Asymptotic Variance of the Output Process of Finite.
1 FARIMA(p,d,q) Model and Application n FARIMA Models -- fractional autoregressive integrated moving average n Generating FARIMA Processes n Traffic Modeling.
Parameter Estimation Problems in Queueing and Related Stochastic Models Yoni Nazarathy School of Mathematics and Physics, The University of Queensland.
1 [3] Jorge Martinez-Bauset, David Garcia-Roger, M a Jose Domenech- Benlloch and Vicent Pla, “ Maximizing the capacity of mobile cellular networks with.
Network Design and Analysis-----Wang Wenjie Queueing System IV: 1 © Graduate University, Chinese academy of Sciences. Network Design and Analysis Wang.
Yoni Nazarathy Gideon Weiss University of Haifa Yoni Nazarathy Gideon Weiss University of Haifa The Asymptotic Variance of the Output Process of Finite.
Matrix Analytic Methods - Some Real Life Applications
Modeling and Analysis of Computer Networks
Pravin Rajamoney CSE-581 Network Technology
CS433 Modeling and Simulation Lecture 12 Queueing Theory Dr. Anis Koubâa 03 May 2008 Al-Imam Mohammad Ibn Saud University.
1 Chapters 8 Overview of Queuing Analysis. Chapter 8 Overview of Queuing Analysis 2 Projected vs. Actual Response Time.
Link Dimensioning for Fractional Brownian Input Chen Jiongze PhD student, Electronic Engineering Department, City University of Hong Kong Supported by.
Yoni Nazarathy Gideon Weiss University of Haifa Yoni Nazarathy Gideon Weiss University of Haifa The Asymptotic Variance Rate of the Departure Process of.
Yoni Nazarathy Gideon Weiss University of Haifa Yoni Nazarathy Gideon Weiss University of Haifa On the Asymptotic Variance Rate of the Output Process of.
June 10, 1999 Discrete Event Simulation - 3 What other subsystems do we need to simulate? Although Packets are responsible for the largest amount of events,
Wireless communications and mobile computing conference, p.p , July 2011.
Multiplicative Wavelet Traffic Model and pathChirp: Efficient Available Bandwidth Estimation Vinay Ribeiro.
© 2006 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice Injecting Realistic Burstiness to.
Self-generated Self-similar Traffic Péter Hága Péter Pollner Gábor Simon István Csabai Gábor Vattay.
Burst Metric In packet-based networks Initial Considerations for IPPM burst metric Tuesday, March 21, 2006.
Maciej Stasiak, Mariusz Głąbowski Arkadiusz Wiśniewski, Piotr Zwierzykowski Model of the Nodes in the Packet Network Chapter 10.
IMA Summer Program on Wireless Communications VoIP over a wired link Phil Fleming Network Advanced Technology Group Motorola, Inc.
Performance Evaluation of Long Range Dependent Queues Performance Evaluation of Long Range Dependent Queues Chen Jiongze Supervisor: Moshe ZukermanCo-Supervisor:
Risk Analysis Workshop April 14, 2004 HT, LRD and MF in teletraffic1 Heavy tails, long memory and multifractals in teletraffic modelling István Maricza.
Eun-Chan Park and Hwangnam Kim Dept. of Information and Communication, Dongguk University ( 南韓東國大學 ) Dept. of Electrical Engineering, Korea University.
Yoni Nazarathy Gideon Weiss University of Haifa Yoni Nazarathy Gideon Weiss University of Haifa The Asymptotic Variance of the Output Process of Finite.
Notices of the AMS, September Internet traffic Standard Poisson models don’t capture long-range correlations. Poisson Measured “bursty” on all time.
1 Interesting Links. On the Self-Similar Nature of Ethernet Traffic Will E. Leland, Walter Willinger and Daniel V. Wilson BELLCORE Murad S. Taqqu BU Analysis.
Queuing Analysis of Tree-Based LRD Traffic Models Vinay J. Ribeiro R. Riedi, M. Crouse, R. Baraniuk.
CS 5253 Workshop 1 MAC Protocol and Traffic Model.
Load Balancing and Data centers
Internet Traffic Modeling
Notices of the AMS, September 1998
مدل سازی ترافیک شبکه استاد راهنما: دکتر کیارش میزانیان
Buffer Management for Shared-Memory ATM Switches
Queueing Theory 2008.
Presentation transcript:

Panel Topic: After Long Range Dependency (LRD) discoveries, what are the lessons learned so far to provide QoS for Internet advanced applications David Lucantoni DLT Consulting, L.L.C.

Markovian is not Poisson  Poisson is Markovian  Markovian >> Poisson MMPP, Markovian Arrival Process (MAP), etc. MMPP, Markovian Arrival Process (MAP), etc.  Typical quote from recent LRD literature: “Prior to the work by Leland, et.al., the Poisson model was the most commonly used model for network traffic” “Prior to the work by Leland, et.al., the Poisson model was the most commonly used model for network traffic” This is incorrect and misleading to students coming into the fieldThis is incorrect and misleading to students coming into the field  Many LRD papers compare their results only to Poisson or at most a 2-state MMPP  Although MAP’s are technically SRD, the exponential tail may not become dominant for some time (medium range dependence?)

Approximating LRD Traffic with an MMPP  2-state on-off processes  Time constants cover several time scales  Superposition of processes is MMPP  Matches correlation lag over 5 orders of magnitude Andersen, A., Nielsen, B., “A Markovian approach for modeling packet traffic with long range dependence”, IEEE Journal on Selected Areas in Communications, vol. 16, no. 5, pp , June, 1998.

Approximating LRD Traffic with an MMPP (Cont’d)  A novel feature of this approach is that the number of time scales is part of fitting procedure  Bellcore data, p.Oct.tl, has 1 million samples  Fit results in an 88-state MMPP  Matches autocorrelation  Matches simulated packet loss Salvador, P., Valadas, R., Pacheco, A, “Multiscale fitting procedure for Markov modulated Poisson processes”, INFOCOM, 2002

Medium Range Dependence Result  Superposition of 60 on-off sources  Exact analysis  Far from Poisson  Exponential tail does not dominate until blocking is Choudhury, Lucantoni, Whitt, “Squeezing the most out of ATM”, IEEE Trans. on Comm, 44, No. 2, , 1996

Known Results for the BMAP/G/1 Queue  Stationary distributions Queue length distribution at arrivals, departures, arbitrary time Queue length distribution at arrivals, departures, arbitrary time Waiting time distributions Waiting time distributions  Transient distributions (given the appropriate initial conditions) P(system empty at time t) P(system empty at time t) Workload at time t Workload at time t Queue length at time t Queue length at time t Delay of n th arrival Delay of n th arrival Queue length at n th departure Queue length at n th departure P(n th departure occurs less than or equal to time x) P(n th departure occurs less than or equal to time x) Lucantoni, D., “The BMAP/G/1 queue: A tutorial”, Models and techniques for Performance Evaluation of Computer and Communications Systems, Editors: L. Donatiello and R. Nelson, Springer Verlag, , 1993

Sample Transient Results - Virtual Delay Distribution  Superposition of four identical two-state MMPP's  Service times are distributed as E 16  Utilization = 0.7 queue length = 0 at time 0 queue length = 0 at time 0 queue length = 32 at time 0 queue length = 32 at time 0 Lucantoni, Choudhury, Whitt, “The transient BMAP/G/1 queue”, Stoch. Mod., 10, No. 1, , 1994

Sample Transient Results - Unstable System  Utilization = 2.0  Queue length = 0 at time 0 Lucantoni, Choudhury, Whitt, “The transient BMAP/G/1 queue”, Stoch. Mod., 10, No. 1, , 1994

State of LRD Performance Models  Several quotes from Park, K., Willinger, W., “Self-similar network traffic: an overview” in Self-Similar Network Traffic and Performance Evaluation, Edited by Kihong Park and Walter Willinger John Wiley & Sons, Inc., 2000

State of LRD Performance Models “A major weakness of many of the [LRD] queueing based results is that they are asymptotic, in one form or another.” “A major weakness of many of the [LRD] queueing based results is that they are asymptotic, in one form or another.” MAP/G/1 results are exact MAP/G/1 results are exact

State of LRD Performance Models “A further drawback of current performance results is that they concentrate on first-order performance measures that relate to (long-term) packet loss rate but less so on second order measures - for example variance of packet loss or delay, generically referred to as jitter - which are of importance in multimedia communication.” “A further drawback of current performance results is that they concentrate on first-order performance measures that relate to (long-term) packet loss rate but less so on second order measures - for example variance of packet loss or delay, generically referred to as jitter - which are of importance in multimedia communication.” Moments as well as distributions can be computed for MAP models Moments as well as distributions can be computed for MAP models

State of LRD Performance Models “Even less is known about transient performance measures, which are more relevant in practice when convergence to long-term steady-state behavior is too slow to be of much value for engineering purposes.” “Even less is known about transient performance measures, which are more relevant in practice when convergence to long-term steady-state behavior is too slow to be of much value for engineering purposes.” As discussed, many transient distributions can be computed exactly for MAP/G/1 As discussed, many transient distributions can be computed exactly for MAP/G/1

VBR Video  LRD community agrees that VBR video is modeled accurately by SRD models Real-time requirements mandate small buffers and large bandwidth Real-time requirements mandate small buffers and large bandwidth Small busy periods mitigate the effect of LRD Small busy periods mitigate the effect of LRD Few arrivals interact in the queueFew arrivals interact in the queue Heyman, D., Lakshman, T., “Long-Range Dependence and Queueing Effects for VBR Video” in Self-Similar Network Traffic and Performance Evaluation, Edited by Kihong Park and Walter Willinger John Wiley & Sons, Inc., 2000

Prediction -  As broadband access becomes more ubiquitous (e.g. cable modem, DSL, etc.) customers expectations will demand better real-time behavior for other types of traffic Consider web browsing. If a user clicks on a link and doesn’t get a response in the expected time the user will click again resulting in wasted requests and bandwidth Consider web browsing. If a user clicks on a link and doesn’t get a response in the expected time the user will click again resulting in wasted requests and bandwidth  It may be necessary to engineer more and more types of traffic with small buffers and higher bandwidth

Prediction (Cont’d)  By the time accurate queueing results become available for LRD traffic models, it might be that most of the traffic in the network will be served like VBR video  Hence LRD would not have any effect on the performance of most traffic

Conclusion  If networks are operated in regions where LRD might have an effect on performance then existing Markovian models can be used for dimensioning  In the future, the nature of customer requirements might force the network to be operated in regions where LRD has no effect