KauNetEm Deterministic Network Emulation in Linux . Per Hurtig, - PowerPoint PPT Presentation

Proceedings of NetDev 1.1: The Technical Conference on Linux Networking (February 10th-12th 2016. Seville, Spain) KauNetEm Deterministic Network Emulation in Linux . Per Hurtig, Johan Garcia February 12, 2016 netdevconf 1.1, February 10-12, Seville, Spain 1 .

Talk outline Proceedings of NetDev 1.1: The Technical Conference on Linux Networking (February 10th-12th 2016. Seville, Spain) 1. Deterministic Network Emulation 2. Use Case Examples 3. KauNetEm System Design 4. Demo 5. Open Issues 6. The Way Forward / Closing Remarks 2 .

Proceedings of NetDev 1.1: The Technical Conference on Linux Networking (February 10th-12th 2016. Seville, Spain) Deterministic Network Emulation .

What? Proceedings of NetDev 1.1: The Technical Conference on Linux Networking (February 10th-12th 2016. Seville, Spain) Deterministic network emulation allows the experimenter not only to generate various emulation effects such as packet loss, rate restrictions or delay, but to apply these emulation effects at precisely controlled places. 4 .

Emulation Setups Proceedings of NetDev 1.1: The Technical Conference on Linux Networking (February 10th-12th 2016. Seville, Spain) 5 .

NetEm Workflow Proceedings of NetDev 1.1: The Technical Conference on Linux Networking (February 10th-12th 2016. Seville, Spain) 6 .

KauNetEm Workflow Proceedings of NetDev 1.1: The Technical Conference on Linux Networking (February 10th-12th 2016. Seville, Spain) 11 .

Data vs. Time driven Proceedings of NetDev 1.1: The Technical Conference on Linux Networking (February 10th-12th 2016. Seville, Spain) 12 .

Emulation Effects Proceedings of NetDev 1.1: The Technical Conference on Linux Networking (February 10th-12th 2016. Seville, Spain) Emulation effect Data-driven Time-driven Packet loss X X Delay X X Rate X X Bit error X - Duplication X X Reordering X - Trigger -X- -?- 13 .

Proceedings of NetDev 1.1: The Technical Conference on Linux Networking (February 10th-12th 2016. Seville, Spain) Use Case Examples .

Why? Proceedings of NetDev 1.1: The Technical Conference on Linux Networking (February 10th-12th 2016. Seville, Spain) • Transport protocol implementation debugging / validation • Functional evaluation of new Transport layer mechanisms • Transport or Application layer performance evaluations • ''Control what you can, and randomize the rest'' 15 .

Random losses Proceedings of NetDev 1.1: The Technical Conference on Linux Networking (February 10th-12th 2016. Seville, Spain) 16 .

Randomly generated loss pattern Proceedings of NetDev 1.1: The Technical Conference on Linux Networking (February 10th-12th 2016. Seville, Spain) 17 .

Increased statistical strength Proceedings of NetDev 1.1: The Technical Conference on Linux Networking (February 10th-12th 2016. Seville, Spain) 18 .

Functional evaluation of Transport layer mechanisms Proceedings of NetDev 1.1: The Technical Conference on Linux Networking (February 10th-12th 2016. Seville, Spain) tail loss behavior of TCP 1800 early_retrans=3 early_retrans=0 1600 1400 transmission time of flow [ms] 1200 1000 800 600 400 200 0 5 10 15 20 position of loss (pkt no.) 19 .

Proceedings of NetDev 1.1: The Technical Conference on Linux Networking (February 10th-12th 2016. Seville, Spain) KauNetEm System Design .

How? Proceedings of NetDev 1.1: The Technical Conference on Linux Networking (February 10th-12th 2016. Seville, Spain) Pattern Creation Pattern/Table Loading User Space Kernel Space Pattern/Table Management Pattern Decoding and Forwarding Loss Duplication Corruption … Traffic NetEm/KauNetEm KauNetEm 21 .

Data representation - Float data Proceedings of NetDev 1.1: The Technical Conference on Linux Networking (February 10th-12th 2016. Seville, Spain) 15 14 13 12 11 10 9 8 7 6 5 4 3 3 2 1 0 F 11 bit mantissa 4 bit exponent F 15 bit run length value F: Flag to indicate if run-length value or float value is encoded • Used for Bandwidth and Delay patterns • 0 - 2 . 047 · 10 0 − 15 [bps, µ s ] (i.e max at ∼ 2Pbps, 2 · 10 9 s) 22 .

Data representation - Float data Proceedings of NetDev 1.1: The Technical Conference on Linux Networking (February 10th-12th 2016. Seville, Spain) 11+4 Float representation error 0.25 0.20 0.15 Percent 0.10 0.05 0.00 2 4 6 8 10 23 .

Data representation - Packet data Proceedings of NetDev 1.1: The Technical Conference on Linux Networking (February 10th-12th 2016. Seville, Spain) 15 14 13 12 11 10 9 8 7 6 5 4 3 3 2 1 0 F 15 bit run length value F: Indicates if packet at current position should be dropped / duplicated • Used for Paket loss and Packet duplication patterns • 0.1% packet loss rate at 100Mbps ⇒ 24 minutes / 16KByte pattern 24 .

Data representation - Integer data Proceedings of NetDev 1.1: The Technical Conference on Linux Networking (February 10th-12th 2016. Seville, Spain) 15 14 13 12 11 10 9 8 7 6 5 4 3 3 2 1 0 F 15 bit integer value F 15 bit run length value F: Flag to indicate if run-length value or float value is encoded • Used for Reordering, Bit errors, and future Trigger patterns • 14 bits to encode bit to flip / reordering distance / trigger value 25 .

User Space Code Extensions Proceedings of NetDev 1.1: The Technical Conference on Linux Networking (February 10th-12th 2016. Seville, Spain) • tc/q_netem.c: 75 lines of code added • Command line parsing • Pattern loading/transfer • patt_gen: ∼ 1400 lines of c code • Value encoding/decoding routines • Pattern creation handling • I/O 26 .

Pattern generation Proceedings of NetDev 1.1: The Technical Conference on Linux Networking (February 10th-12th 2016. Seville, Spain) patt_gen -pkt|-ber|-del|-bw|-reo|-dup|-trig -s <size> [-o <outfilename>] <position-values>|-f infilename <position-values>: A comma separated list of positions and values, or loss/duplication positions. <size>: Specifies the length of the generated pattern. The unit is packets or milliseconds. For packet loss patterns: patt_gen -pkt -rand -s size [-r random_seed] <PVAL> patt_gen -pkt -ge -s size [-r random_seed] <good_rate> <bad_rate> <good_tran_prob> <bad_tran_prob> patt_gen -pkt -int -s size <interval-list>|-f <infilename> 27 .

Packet loss probability vs Actual loss rate Proceedings of NetDev 1.1: The Technical Conference on Linux Networking (February 10th-12th 2016. Seville, Spain) There is a difference between a 0.2% packet loss probability and an actual 0.2% achieved loss rate. Example: 0.2% loss for a 1000 packet flow: PVAL = 0.002 PVAL = 2 0.5 1.0 Probability mass Probability mass 0.4 0.8 0.6 0.3 0.2 0.4 0.1 0.2 0.0 0.0 0 2 4 6 8 10 0 2 4 6 8 10 28 .

Kernel Space Code Extensions Proceedings of NetDev 1.1: The Technical Conference on Linux Networking (February 10th-12th 2016. Seville, Spain) • pkt_sched.h: 12 lines (2 variables, 10 defines) • sch_netem.c: ~400 lines • Management and traversal of patterns • pattern forwarding • value decoding, • hrtimer management • Pattern effect invocation 29 .

Proceedings of NetDev 1.1: The Technical Conference on Linux Networking (February 10th-12th 2016. Seville, Spain) Demo .

Packet Loss Proceedings of NetDev 1.1: The Technical Conference on Linux Networking (February 10th-12th 2016. Seville, Spain) 31 .

Duplication Proceedings of NetDev 1.1: The Technical Conference on Linux Networking (February 10th-12th 2016. Seville, Spain) 32 .

Proceedings of NetDev 1.1: The Technical Conference on Linux Networking (February 10th-12th 2016. Seville, Spain) Open Issues .

Open design considerations Proceedings of NetDev 1.1: The Technical Conference on Linux Networking (February 10th-12th 2016. Seville, Spain) • What should happen at the end of a pattern? • Just end pattern? • Wraparound? • Append new pattern? • Behavior for packet being sent at low throughput when time-driven increase of throughput happens? • Send remaining bits at original low rate? • Send remaining bits at new higher rate? • Need for additional patterns? • Trigger patterns will be added • Any more? • Fold patt_gen code into tc? • Split delay patterns to delay_reorder and delay_fifo? 34 .

KauNetEm Deterministic Network Emulation in Linux . Per Hurtig, - PowerPoint PPT Presentation

Proceedings of NetDev 1.1: The Technical Conference on Linux Networking (February 10th-12th 2016. Seville, Spain) KauNetEm Deterministic Network Emulation in Linux . Per Hurtig, Johan Garcia February 12, 2016 netdevconf 1.1, February 10-12,

Bayesian Emulation and Calibration of a Dynamic Epidemic Model for H1N1 Influenza Marian Farah 1

Using Reconfigurable Logic Using Reconfigurable Logic to Simulate Computer Systems Derek Chiou

Mobile Applications Alberto Garca Estvez University of Alcala Niklas Carlsson Linkping

FPGA-Based Circuit Model Emulation of Quantum Algorithms Mahdi Aminian, Mehdi Saeedi, Morteza

The microkernel OS Escape Nils Asmussen FOSDEM14 1 / 25 Introduction Tasks Memory VFS

Crash Course Entrepreneurship Crash Course Escape from Corporate [Case Study] Who wants

Escaping chroot jails Why? Chroot jails come up in writing exploits, CtF competitions, etc.

Empir iric ical l Aspects of Plu Pluralit lity Ele lectio ions David R. M. Thompson, Omer

Justifying the state: Escaping the State of Nature Trading Liberty for Protection Hobbes

Extended Example (Iterative Refinement) A maze consists of rooms and doors: An door is either

Iteration in tracts James Waterman Department of Mathematics and Statistics The Open University

01-09-2014 Outline Preconditions Outcome: What are you supposed to learn? The framework and

Practical Philosophy Political and Philosophy of Ethics Social Aesthetics Law Philosophy

W hat is Robotics Robotics is the science that studies robots and the CY 02CFI C technology

Final exam Patrick Loiseau, Michela Chessa Game Theory, Fall 2013 2 hours, no document allowed

Evolutionary Games with Time Constraints Vlastimil Krivan Biology Center and Faculty of Science

Models of Language Evolution Evolutionary game theory & the evolution of meaning Michael

Introduction to Game Theory Review for the Final Exam Dana Nau University of Maryland Updated

Conflicting Interests and Language Change GURT 2014 Christopher Ahern & Robin Clark

Dynamics on Games: Simulation-Based Techniques and Applications to Routing Benjamin Monmege

Evolving Game Playing Evolving Game Playing Strategies (4.4.3) Strategies (4.4.3) Darren

Game Theoretical Approaches to the Handling Road Traffic Ayantha Randika Overview What is

The Evolution of Cooperation Through Imitation David K. Levine and Wolfgang Pesendorfer 1

Bismark Owusu Tawiah University of LAquila -Italy OVERVIEW Introduction Competitive

KauNetEm Deterministic Network Emulation in Linux . Per Hurtig, - PowerPoint PPT Presentation

Proceedings of NetDev 1.1: The Technical Conference on Linux Networking (February 10th-12th 2016. Seville, Spain) KauNetEm Deterministic Network Emulation in Linux . Per Hurtig, Johan Garcia February 12, 2016 netdevconf 1.1, February 10-12,

Bayesian Emulation and Calibration of a Dynamic Epidemic Model for H1N1 Influenza Marian Farah 1

Using Reconfigurable Logic Using Reconfigurable Logic to Simulate Computer Systems Derek Chiou

Mobile Applications Alberto Garca Estvez University of Alcala Niklas Carlsson Linkping

FPGA-Based Circuit Model Emulation of Quantum Algorithms Mahdi Aminian, Mehdi Saeedi, Morteza

The microkernel OS Escape Nils Asmussen FOSDEM14 1 / 25 Introduction Tasks Memory VFS

Crash Course Entrepreneurship Crash Course Escape from Corporate [Case Study] Who wants

Escaping chroot jails Why? Chroot jails come up in writing exploits, CtF competitions, etc.

Empir iric ical l Aspects of Plu Pluralit lity Ele lectio ions David R. M. Thompson, Omer

Justifying the state: Escaping the State of Nature Trading Liberty for Protection Hobbes

Extended Example (Iterative Refinement) A maze consists of rooms and doors: An door is either

Iteration in tracts James Waterman Department of Mathematics and Statistics The Open University

01-09-2014 Outline Preconditions Outcome: What are you supposed to learn? The framework and

Practical Philosophy Political and Philosophy of Ethics Social Aesthetics Law Philosophy

W hat is Robotics Robotics is the science that studies robots and the CY 02CFI C technology

Final exam Patrick Loiseau, Michela Chessa Game Theory, Fall 2013 2 hours, no document allowed

Evolutionary Games with Time Constraints Vlastimil Krivan Biology Center and Faculty of Science

Models of Language Evolution Evolutionary game theory &amp; the evolution of meaning Michael

Introduction to Game Theory Review for the Final Exam Dana Nau University of Maryland Updated

Conflicting Interests and Language Change GURT 2014 Christopher Ahern &amp; Robin Clark

Dynamics on Games: Simulation-Based Techniques and Applications to Routing Benjamin Monmege

Evolving Game Playing Evolving Game Playing Strategies (4.4.3) Strategies (4.4.3) Darren

Game Theoretical Approaches to the Handling Road Traffic Ayantha Randika Overview What is

The Evolution of Cooperation Through Imitation David K. Levine and Wolfgang Pesendorfer 1

Bismark Owusu Tawiah University of LAquila -Italy OVERVIEW Introduction Competitive

Models of Language Evolution Evolutionary game theory & the evolution of meaning Michael

Conflicting Interests and Language Change GURT 2014 Christopher Ahern & Robin Clark