Network Tomography Based on Flow Level Measurements Dogu Arifler, Gustavo de Veciana, and Brian L. Evans The University of Texas at Austin IEEE International.

Slides:



Advertisements
Similar presentations
Multicast congestion control on many-to- many videoconferencing Xuan Zhang Network Research Center Tsinghua University, China.
Advertisements

Traffic Dynamics at a Commercial Backbone POP Nina Taft Sprint ATL Co-authors: Supratik Bhattacharyya, Jorjeta Jetcheva, Christophe Diot.
Ningning HuCarnegie Mellon University1 Optimizing Network Performance In Replicated Hosting Peter Steenkiste (CMU) with Ningning Hu (CMU), Oliver Spatscheck.
G-RCA: A Generic Root Cause Analysis Platform for Service Quality Management in Large IP Networks He Yan, Lee Breslau, Zihui Ge, Dan Massey, Dan Pei, Jennifer.
Doc.: IEEE /0604r1 Submission May 2014 Slide 1 Modeling and Evaluating Variable Bit rate Video Steaming for ax Date: Authors:
Improving TCP Performance over Mobile Ad Hoc Networks by Exploiting Cross- Layer Information Awareness Xin Yu Department Of Computer Science New York University,
Fast Pattern-Based Throughput Prediction for TCP Bulk Transfers Tsung-i (Mark) Huang Jaspal Subhlok University of Houston GAN ’ 05 / May 10, 2005.
PROMISE: Peer-to-Peer Media Streaming Using CollectCast Mohamed Hafeeda, Ahsan Habib et al. Presented By: Abhishek Gupta.
TCP Stability and Resource Allocation: Part II. Issues with TCP Round-trip bias Instability under large bandwidth-delay product Transient performance.
Source-Adaptive Multilayered Multicast Algorithms for Real- Time Video Distribution Brett J. Vickers, Celio Albuquerque, and Tatsuya Suda IEEE/ACM Transactions.
Traffic Engineering With Traditional IP Routing Protocols
Internet Traffic Patterns Learning outcomes –Be aware of how information is transmitted on the Internet –Understand the concept of Internet traffic –Identify.
Server-based Inference of Internet Performance V. N. Padmanabhan, L. Qiu, and H. Wang.
Factor Analysis There are two main types of factor analysis:
Low Delay Marking for TCP in Wireless Ad Hoc Networks Choong-Soo Lee, Mingzhe Li Emmanuel Agu, Mark Claypool, Robert Kinicki Worcester Polytechnic Institute.
Multi-Scale Analysis for Network Traffic Prediction and Anomaly Detection Ling Huang Joint work with Anthony Joseph and Nina Taft January, 2005.
Network Traffic Measurement and Modeling CSCI 780, Fall 2005.
Available bandwidth measurement as simple as running wget D. Antoniades, M. Athanatos, A. Papadogiannakis, P. Markatos Institute of Computer Science (ICS),
1 TCP-LP: A Distributed Algorithm for Low Priority Data Transfer Aleksandar Kuzmanovic, Edward W. Knightly Department of Electrical and Computer Engineering.
1 End-to-End Detection of Shared Bottlenecks Sridhar Machiraju and Weidong Cui Sahara Winter Retreat 2003.
Copyright © 2005 Department of Computer Science CPSC 641 Winter Network Traffic Measurement A focus of networking research for 20+ years Collect.
1 Emulating AQM from End Hosts Presenters: Syed Zaidi Ivor Rodrigues.
Rethinking Internet Traffic Management: From Multiple Decompositions to a Practical Protocol Jiayue He Princeton University Joint work with Martin Suchara,
Estimating Congestion in TCP Traffic Stephan Bohacek and Boris Rozovskii University of Southern California Objective: Develop stochastic model of TCP Necessary.
Wide Web Load Balancing Algorithm Design Yingfang Zhang.
Multipath Protocol for Delay-Sensitive Traffic Jennifer Rexford Princeton University Joint work with Umar Javed, Martin Suchara, and Jiayue He
1 A State Feedback Control Approach to Stabilizing Queues for ECN- Enabled TCP Connections Yuan Gao and Jennifer Hou IEEE INFOCOM 2003, San Francisco,
MATE: MPLS Adaptive Traffic Engineering Anwar Elwalid, et. al. IEEE INFOCOM 2001.
Tomo-gravity Yin ZhangMatthew Roughan Nick DuffieldAlbert Greenberg “A Northern NJ Research Lab” ACM.
A User Experience-based Cloud Service Redeployment Mechanism KANG Yu.
Network Monitoring School of Electronics and Information Kyung Hee University. Choong Seon HONG Selected from ICAT 2003 Material of James W. K. Hong.
Lect3..ppt - 09/12/04 CIS 4100 Systems Performance and Evaluation Lecture 3 by Zornitza Genova Prodanoff.
Distributed Quality-of-Service Routing of Best Constrained Shortest Paths. Abdelhamid MELLOUK, Said HOCEINI, Farid BAGUENINE, Mustapha CHEURFA Computers.
Server Load Balancing. Introduction Why is load balancing of servers needed? If there is only one web server responding to all the incoming HTTP requests.
A Statistical Anomaly Detection Technique based on Three Different Network Features Yuji Waizumi Tohoku Univ.
Tiziana Ferrari Quality of Service Support in Packet Networks1 Quality of Service Support in Packet Networks Tiziana Ferrari Italian.
CS540/TE630 Computer Network Architecture Spring 2009 Tu/Th 10:30am-Noon Sue Moon.
Computer Networks Performance Metrics. Performance Metrics Outline Generic Performance Metrics Network performance Measures Components of Hop and End-to-End.
UDT: UDP based Data Transfer Protocol, Results, and Implementation Experiences Yunhong Gu & Robert Grossman Laboratory for Advanced Computing / Univ. of.
ACN: RED paper1 Random Early Detection Gateways for Congestion Avoidance Sally Floyd and Van Jacobson, IEEE Transactions on Networking, Vol.1, No. 4, (Aug.
Measurement and Modeling of Packet Loss in the Internet Maya Yajnik.
Hung X. Nguyen and Matthew Roughan The University of Adelaide, Australia SAIL: Statistically Accurate Internet Loss Measurements.
Chapter 12 Transmission Control Protocol (TCP)
Network-Coding Multicast Networks With QoS Guarantees Yuanzhe Xuan and Chin-Tau Lea, Senior Member, IEEE IEEE/ACM TRANSACTIONS ON NETWORKING, VOL. 19,
Congestion Control in CSMA-Based Networks with Inconsistent Channel State V. Gambiroza and E. Knightly Rice Networks Group
ﺑﺴﻢﺍﷲﺍﻠﺭﺣﻣﻥﺍﻠﺭﺣﻳﻡ. Group Members Nadia Malik01 Malik Fawad03.
Jennifer Rexford Princeton University MW 11:00am-12:20pm Measurement COS 597E: Software Defined Networking.
Wireless communications and mobile computing conference, p.p , July 2011.
Trajectory Sampling for Direct Traffic Oberservation N.G. Duffield and Matthias Grossglauser IEEE/ACM Transactions on Networking, Vol. 9, No. 3 June 2001.
Deadline-based Resource Management for Information- Centric Networks Somaya Arianfar, Pasi Sarolahti, Jörg Ott Aalto University, Department of Communications.
T. S. Eugene Ngeugeneng at cs.rice.edu Rice University1 COMP/ELEC 429/556 Introduction to Computer Networks Principles of Congestion Control Some slides.
1 Analysis of a window-based flow control mechanism based on TCP Vegas in heterogeneous network environment Hiroyuki Ohsaki Cybermedia Center, Osaka University,
Network Tomography Based on Flow Level Measurements Dogu Arifler Ph.D. Defense Committee Members: Prof. Ross Baldick Prof. Melba M. Crawford Prof. Gustavo.
We used ns-2 network simulator [5] to evaluate RED-DT and compare its performance to RED [1], FRED [2], LQD [3], and CHOKe [4]. All simulation scenarios.
정하경 MMLAB Fundamentals of Internet Measurement: a Tutorial Nevil Brownlee, Chris Lossley, “Fundamentals of Internet Measurement: a Tutorial,” CMG journal.
ANOVA, Regression and Multiple Regression March
Development of a QoE Model Himadeepa Karlapudi 03/07/03.
Advanced Statistics Factor Analysis, I. Introduction Factor analysis is a statistical technique about the relation between: (a)observed variables (X i.
Feature Extraction 主講人:虞台文. Content Principal Component Analysis (PCA) PCA Calculation — for Fewer-Sample Case Factor Analysis Fisher’s Linear Discriminant.
Performance Limitations of ADSL Users: A Case Study Matti Siekkinen, University of Oslo Denis Collange, France Télécom R&D Guillaume Urvoy-Keller, Ernst.
Principal Component Analysis
1 Internet Traffic Measurement and Modeling Carey Williamson Department of Computer Science University of Calgary.
1 Network Tomography Using Passive End-to-End Measurements Lili Qiu Joint work with Venkata N. Padmanabhan and Helen J. Wang.
TCP/IP1 Address Resolution Protocol Internet uses IP address to recognize a computer. But IP address needs to be translated to physical address (NIC).
Chapter 14 EXPLORATORY FACTOR ANALYSIS. Exploratory Factor Analysis  Statistical technique for dealing with multiple variables  Many variables are reduced.
1 Stochastic Ordering for Internet Congestion Control Han Cai, Do Young Eun, Sangtae Ha, Injong Rhee, and Lisong Xu PFLDnet 2007 February 7, 2007.
Estimating standard error using bootstrap
Fast Pattern-Based Throughput Prediction for TCP Bulk Transfers
Dogu Arifler and Brian L. Evans
Principal Component Analysis
Presentation transcript:

Network Tomography Based on Flow Level Measurements Dogu Arifler, Gustavo de Veciana, and Brian L. Evans The University of Texas at Austin IEEE International Conference on Acoustics, Speech, and Signal Processing Montréal, Canada, May 18,

2 Outline Introduction Motivation for inferring network resource sharing Flow level measurements Methodology for inferring network resource sharing Sampling of flow class throughput processes Dimensionality reduction Validation with measured data TCP measurements from UT Austin’s border router Statistical accuracy of estimates Conclusion

3 Inference of congested resource sharing Motivation: Network managers need information about resource sharing in other networks to better plan for services and diagnose performance problems Internet service providers need to diagnose configuration errors and link failures in peer networks Content providers need to balance workload and plan cache placement Problem: In general, properties of networks outside one’s administrative domain are unknown Little or no information on routing, topology, or link utilizations Solution: Network tomography Inferring characteristics of networks from available network traffic measurements

4 Network tomography Previous work based on packet level measurements Correlation of end-to-end packet losses and delays [Rubenstein, Kurose & Towsley, 2002] Inspection of arrival order of packets using probe packets [Rabbat, Nowak & Coates, 2002] Data intensive to collect and store each packet Complex to analyze: high variability over different time scales [Feldmann, Gilbert, Huang & Willinger, 2002] Propose to use flow level measurements A flow is a sequence of packets associated with a given instance of an application [IETF RFC #2722, 1999] Packets composing a flow correspond to transfer of a Web page, a file, an message, etc. Passive flow level measurements available at local site

5 Flow level measurements Flow records Provide summary information Easier to collect and store Collected by networking equipment (e.g. Cisco NetFlow, sFlow, Argus) Flow records contain Source/destination IP addresses, port numbers, number of packets and bytes in flow, and start and end time of flow ~80% of Internet flows are TCP flows [ Data warehouse RecordsMonitored link packets of a flow timeout time start time end time response time identifier 1 identifier 2

6 TCP flows TCP adapts its data transmission rate to available network capacity Congested link bandwidth sharing is roughly fair for flows that have similar packet loss rates and roundtrip times Correlated link bandwidth allocation among flows results in correlated flow performance measures TCP flow performance measure: perceived throughput Amount of data in bytes (flow size) divided by response time Premise: Throughputs of TCP flows that temporally overlap at a congested resource are correlated time available capacity flow 1 flow 2

7 Throughput of a flow class Flow class is a collection of flow records that have a common identifier, e.g. source/destination address How can one infer which flow classes share resources? Correlate flow class throughput processes given by time... class 2 class 1 Flow records collected at a measurement site

8 Which flow class throughput samples can be used to capture flow class throughput correlations? Construct correlation matrix R of pairwise correlations Estimate throughput correlation between class pairs by using class throughput samples at times when flow class pair is active N(T) number of discrete intervals over which c i and c j are active Conditional sampling of random processes time consider red and blue classes Example activity of flow classes for discrete time index n

9 Exploratory factor analysis Correlation structure captured by few latent factors Orthogonal factor model p flow classes and m factors where m ≤ p Λ ij are loadings (or weights) of each factor F i on a variable Estimate Λ and using principal components analysis m determined by H. F. Kaiser’s rule [1960] : Principal components whose variances are greater than 1 are significant factors

10 Inference of resource sharing Class 1 Class 2 Class 3 Class 4 Class 5 Factor 1Factor 2 Classes 1, 2 and 5 share one resource Classes 3 & 4 share another resource Consider five flow classes with two significant factors identified Factor loading with largest magnitude in each row is boxed Source Destination Paper validates approach using known distributions of flow sizes and flow arrivals for two topologies

11 Measured data: preprocessing Two NetFlow datasets from UT Austin’s border router Assume that traffic is stationary over one-hour periods Choose two incoming flow classes that are very likely to experience congestion at the server Select IP addresses associated with AOL and HotMail Divide each class into two: AOL1, AOL2 and HotMail1, HotMail2 Filter flow records based on Packets: Discard flows consisting of only 1 packet Duration: Discard flows with duration shorter than 1 second Size: Discard flows with sizes 64 kB Collection datePeriodTCP records Dataset200211/06/200212:58-2:07 PM5,173,385 Dataset200401/21/200412:58-1:26 PM4,440,697

12 Measured data: component variances Parent class (AOL and HotMail) throughput correlation is for Dataset2002 and 0.05 for Dataset % bootstrap confidence intervals of variances of principal components of 4 classes AOL1, AOL2, HotMail1, and HotMail2 given below  2 significant factors have explanatory power of 72% for Dataset2002 and 63% for Dataset2004 Principal component Dataset % confidence interval Dataset % confidence interval 1(1.5457, )(1.3646, ) 2(1.0861, )(1.0237, ) 3(0.7058, )(0.8230, ) 4(0.2194, )(0.5413, )

13 Measured data: factor loadings Based on 2 significant factors, determine factor loadings Rotated factor loading estimates Rows correspond to classes Columns correspond to shared infrastructure Estimate 95% bootstrap confidence intervals for loadings to establish accuracy † With 95% confidence, we can identify which flow classes share infrastructure Dataset2002Dataset2004 AOL1 AOL2 HotMail1 Hotmail2 AOL1 AOL2 HotMail1 Hotmail2 † Dogu Arifler, Network Tomography Based on Flow Level Measurements, Ph.D. Dissertation, 2004.

14 Conclusion Contributions Application of a structural analysis technique, factor analysis, to explore network properties Methodology for inferring resource sharing Use of bootstrap methods to make inferential statements about resource sharing Possible applications Network monitoring and root cause analysis of poor performance Problem diagnosis and off-line evaluation of congestion status of networks Route configuration by Internet service providers

15 Backup slides

16 Flow level performance of elastic traffic Elastic traffic can tolerate rate variations This implies that a closed-loop control, such as TCP, can be applied end-to-end on flows Additive increase, multiplicative decrease congestion avoidance algorithm of TCP The transmission rate increases linearly in the absence of packet loss, and is halved when there is packet loss For a given RTT and loss rate p, flow throughput is: Also, this relates p to throughput However, y(p) depends on number of flows in progress Packet level dynamics is determined by flow level dynamics

17 Notes on processor sharing When there are n customers in the system, each receive service at a rate 1/n sec/sec All customers are sharing the capacity equally Two abstractions: Customers are given the full capacity on a part-time basis Customers are given a fractional-capacity on a full-time basis Why does TCP realize processor sharing? When there are n flows in the single-bottleneck system, the protocol tends to share bandwidth roughly equally among flows (for flows with similar RTTs and packet loss rates). This is processor sharing! More generally, TCP’s additive-increase/multiplicative-decrease (AIMD) achieves fair sharing [Massoulie and Roberts, 2002]

18 Notes on factor analysis Factor analysis vs. principal component analysis (PCA) In factor analysis, primary goal is to explain correlations between variables (off-diagonal elements of covariance/correlation matrix) In PCA, primary goal is to explain variance (diagonal elements of covariance/correlation matrix) PCA is usually used to find initial estimates of loadings Another related method: independent component analysis; looks at higher order moments How do temporal correlations within a class’ throughput affect factor analysis? Ignore serial correlations when the interest is descriptive or exploratory in nature Successfully applied to econometric time series, biometric time series, etc. See e.g., Basilevsky 1993 or Jolliffe 2002

19 Confidence intervals for loadings

20 Interaction of coupled traffic Consider a “linear” network to evaluate the effect of interactions of coupled network traffic Can throughputs of two flow classes that do not share a link be correlated due to interactions through another flow class? Results of fluid simulations show that degree of correlation between throughputs of classes not sharing a link is negligible file server 3 10 Mbps LANs with 10 workstations file server 1 file server 2

21 Interaction of coupled traffic: an example Consider the “linear” network below Discard flows with sizes 32 kB Based on 2 significant factors, determine factor loadings Rotated factor loading estimates Rows correspond to classes Columns correspond to shared links file server 3 10 Mbps LANs with 10 workstations file server 1 file server 2 80% Background traffic utilizes 20% of bottleneck links (20%) (40%)

22 The bootstrap Validation with real data is extremely difficult! Unlike controlled simulations, we do not know routing information We would like to be able to make inferential statements Estimate 95% confidence intervals for eigenvalues and loadings Modify Kaiser’s rule for selecting significant eigenvalues The bootstrap, a computer-based method, can be used to compute confidence intervals [Efron and Tibshirani, 1993] From data at hand, construct empirical distribution and generate many realizations No distributional assumptions on data required Applicable to any statistic, s(X), simple or complicated (B independent replications) samples of size n

23 Principal component method Determine m “significant” eigenvalues of R using Kaiser’s rule [Kaiser, 1960] Variances of factors are given by eigenvalues eigenvalue variance of a normalized variable … m significant eigenvalues Use spectral decomposition on R to estimate Λ and Eigenvalue-eigenvector pairs ( i, ξ i ), 1 ≤ i ≤ p where

24 Methodology for inferring resource sharing 1.Define the flow classes of interest, C 2.Set flow filtering thresholds for packets, duration, and size 3.Determine flows F that satisfy the filtering criteria 4.Compute flow class throughputs at discretized times 5.Through conditional sampling, estimate pairwise correlations 6.Find number of factors m using eigenvalues of the correlation matrix and modified Kaiser's rule 7.Perform exploratory factor analysis based on m factors 8.Rotate factor loadings using varimax rotation 9.Determine which flow classes have the largest loading on a given factor: Inference of shared congested resources

25 Summary of methodology Flow filtering Bootstrap Exploratory factor analysis Conditional sampling Network tomography