Ge#ng

Started
(1) Copy to your hard disk from a USB Key or DVD:
Copy needed les (VirtualBox, terminal, possibly an X server) for your plaGorm (Win/Mac/Linux) Copy Java 6 and Eclipse for your plaGorm, if you want to use Java Copy VM image: OpenFlowTutorial-101311.zip Pass on the DVD or USB key to someone else!

(2) Unzip OpenFlowTutorial-101311.zip

(3) Point browser to instruc\ons:

h]p://www.openow.org/wk/index.php/OpenFlow_Tutorial (note the underscore)

You should NOT need to download any large les spare the WiFi!
1

Tutorial 1: SDN for Engineers

part of the the Open Networking Summit Santa Clara Marrio] April 16, 2012

Brandon Heller, Rob Sherwood, David Erickson, Hideyuki Shimonishi, Srini Seetharaman, Murphy McCauley
only possible help from all the people listed on the next few pages

Ge#ng Started
(1) Copy to your hard disk from a USB Key or DVD:
Copy needed les (VirtualBox, terminal, possibly an X server) for your plaGorm (Win/Mac/Linux) Copy Java 6 and Eclipse for your plaGorm, if you want to use Java Copy VM image: OpenFlowTutorial-101311.zip Pass on the DVD or USB key to someone else!

(2) Unzip OpenFlowTutorial-101311.zip

(3) Point browser to instruc\ons:

h]p://www.openow.org/wk/index.php/OpenFlow_Tutorial (note the underscore)

You should NOT need to download any large les spare the WiFi!
3

Welcome

Whos in this room?

Whos in this room?

>= 3: HP >= 19 Cisco >= 15 Juniper >= 7 Huawei >= 6 Brocade >= 4 ZAO >= 3 Google >= 3 Plus >= 60 companies w/2 or less..
6

Whos in this room?

Engineer Senior Engineer Principal Systems Engineer Dis\nguished Engineer Senior Dis\nguished Engineer Senior Technical Marke\ng Engineer

Whos in this room?

Student Graduate Student PhD / Doctor Senior Research Scien\st Professor

Whos in this room?

CTO VP Founder Solu\ons Architect Cloud Architect Chief Tech Expert Chief
9

Goals

10

Overall Goal: think cri\cally about SDN

11

Goals of this Tutorial

By the end, everyone should know:
Knowledge about SDN/OpenFlow
what these are how they relate whats available now where its going how its used

SDN and You

Have fun

how you can use it how you can build on top of whats available How you can build something completely new

12

Agenda
Time 9:00-10:00 10:00-10:30 10:45-12:00 12:00-1:00 1:00-3:00 3:15-4:15 4:20-5:00 DescripEon OpenFlow/SDN Introduc\on The SDN Stack, Part 1: Switches & Controllers The SDN Stack, Part 2: Virtualiza\on & SDN Applica\ons Lunch Controllers + Q&A Hands-On SDN Deployment Experiences and Wrap-up
13

SDN + OpenFlow History

14

Why OpenFlow? [Historical mo\va\ons, for researchers]

15

The Networking Industry (2007)

Rou\ng, management, mobility management, access control, VPNs,
Feature Feature

Opera\ng System Specialized Packet Forwarding Hardware

Million of lines of source code

5400 RFCs

Barrier to entry

Billions of gates Complex

Power Hungry

Many complex func\ons baked into the infrastructure

An industry with a mainframe-mentality Li]le ability for non-telco network operators to get what they want Func\onality dened by standards, put in hardware, deployed on nodes

OSPF, BGP, mul.cast, dieren.ated services, Trac Engineering, NAT, rewalls, MPLS, redundant layers,

16

Ossica\on

Source: http://www.merriam-webster.com/
17

Research Stagna\on (circa 2007): Faster networks but not beFer networks
Lots of deployed innova\on in other areas
OS: lesystems, schedulers, virtualiza\on DS: DHTs, CDNs, MapReduce Compilers: JITs, new language paradigms

Networks are largely the same as years ago

Ethernet, IP, WiFi

Rate of change of the network seems slower in comparison

Need be]er tools and abstrac\ons to demonstrate and deploy
18

Another problem: Closed Systems (Vendor Hardware)

Cant extend Stuck with interfaces (CLI, SNMP, etc) Hard to meaningfully extend Hard to meaningfully collaborate Vendors star\ng to open up, but not usefully

19

Open Systems
Performance Scale Fidelity Simula\on Emula\on Soyware Switches NetFPGA Network Processors Vendor Switches medium medium poor high high high medium low low low medium high Real User Trac? no no yes yes yes yes Complexity medium medium medium high high low Open yes yes yes yes yes no

gap in the tool space none have all the desired a]ributes!

20

Ethane, a precursor to OpenFlow

Centralized, reac\ve, per-ow control

Controller

Flow Switch

Flow Switch

Flow Switch

Host A
Flow Switch

Host B

See Ethane SIGCOMM 2007 paper for details

21

Ethanes dumb, simple switches might be more broadly useful

22

OpenFlow: a pragma\c compromise

+ Speed, scale, delity of vendor hardware + Flexibility and control of soyware and simula\on + Vendors dont need to expose implementa\on + Leverages hardware inside most switches today (ACL tables) - Least-common-denominator interface may prevent using all hardware features - Limited table sizes - Switches not designed for this - New failure modes to understand
23

How does OpenFlow work?

24

Ethernet Switch

25

26

OpenFlow Protocol (SSL/TCP)

27

OpenFlow Example

Controller
PC

Soyware Layer

OpenFlow Client
Flow Table

Hardware Layer

MAC src
*

MAC IP dst Src

* *

IP Dst
5.6.7.8

TCP TCP Ac\on sport dport

* * port 1

port 1

port 2

port 3

port 4

5.6.7.8

1.2.3.4

28

OpenFlow Basics
Flow Table Entries
Rule Ac\on Stats

Packet + byte counters 1. Forward packet to zero or more ports 2. Encapsulate and forward to controller 3. Send to normal processing pipeline 4. Modify Fields 5. Any extensions you add!
Switch VLAN VLAN MAC MAC Port ID pcp src dst Eth type IP Src IP Dst L4 L4 IP IP ToS Prot sport dport

+ mask what elds to match

29

Examples
Switching
Switch MAC Port src * * MAC Eth dst type 00:1f:.. * VLAN IP ID Src * * IP Dst * IP Prot * TCP TCP Ac\on sport dport * * port6

Flow Switching
Switch MAC Port src MAC Eth dst type VLAN IP ID Src IP Dst IP Prot TCP TCP Ac\on sport dport 17264 80 port6

port3 00:20.. 00:1f.. 0800 vlan1 1.2.3.4 5.6.7.8 4

Firewall
Switch MAC Port src * * * MAC Eth dst type * VLAN IP ID Src * * IP Dst * IP Prot * TCP TCP Ac\on sport dport * 22 drop
30

Examples
Rou\ng
Switch MAC Port src * * * MAC Eth dst type * VLAN IP ID Src * * IP Dst IP Prot TCP TCP Ac\on sport dport * * port6

5.6.7.8 *

VLAN Switching
Switch MAC Port src * * MAC Eth dst type 00:1f.. * VLAN IP ID Src vlan1 * IP Dst * IP Prot * TCP TCP Ac\on sport dport port6, port7, * * port9

31

OpenFlow is not enough.

32

OpenFlow is not enough

Adds the ability to modify, experiment But s\ll harder than it should be to add features to a network Eec\vely assembly programming or an ISA

[OpenFlow is just a forwarding table management protocol]

33

Its hard to add a feature to a network

Its not just that we lack access to line-rate forwarding that we can control Fully distributed algorithms are hard, especially when dened at the protocol level Your protocol must implement its own mechanisms Must work on constrained and heterogeneous resources

This is where Software-Defined Networking comes in: need a control plane abstraction too

34

From OpenFlow to SDN

35

36

SDN is actually much older than 09

Key ideas present in some form in 4D, RCP, Sane, Ethane,

37

What is SDN, opt. 1

The McKeown View: Refactoring Func\onality
38

Dene SDN by its placement of func\onality.

39

Closed Boxes, Fully Distributed Protocols

Today

Closed
App App App

Opera\ng System Specialized Packet Forwarding Hardware

Ap p Ap p Ap p

Opera\ng System Specialized Packet Forwarding Hardware

App App App

App

App

App

Opera\ng System Specialized Packet Forwarding Hardware Opera\ng System Specialized Packet Forwarding Hardware

App

App

App

Opera\ng System Specialized Packet Forwarding Hardware

40

Soyware Dened Networking approach to open it

App App App
Network Opera\ng System

App

App

App

Opera\ng System Specialized Packet Forwarding Hardware

Ap p Ap p Ap p

Opera\ng System Specialized Packet Forwarding Hardware

App App App

App

App

App

Opera\ng System Specialized Packet Forwarding Hardware Opera\ng System Specialized Packet Forwarding Hardware

App

App

App

Opera\ng System Specialized Packet Forwarding Hardware

41

The Soyware-dened Network

3. Well-dened open API 2. At least one good opera\ng system Extensible, possibly open-source

App

App

App

Network Opera\ng System

1. Open interface to hardware

Simple Packet Forwarding Hardware

Simple Packet Forwarding Hardware Simple Packet Forwarding Hardware

Simple Packet Forwarding Hardware Simple Packet Forwarding Hardware

42

Isolated slices

Many opera\ng systems, or Many versions

App App App App App App

App

App

Network Opera\ng System 1

Network Opera\ng System 2

Network Opera\ng System 3

Network Opera\ng System 4

Open interface to hardware

Virtualiza\on or Slicing Layer

Open interface to hardware
Simple Packet Forwarding Hardware Simple Packet Forwarding Hardware Simple Packet Forwarding Hardware Simple Packet Forwarding Hardware

Simple Packet Forwarding Hardware

43

What is SDN, opt. 2

The Shenker View: Redening Abstrac\ons
44

Dene SDN by the abstrac\ons it provides to soyware (and people wri\ng it).
45

The Shenker View

Sco] Shenker has a killer presenta\on
(Keynote at last ONS) You should watch this h]p://www.slideshare.net/mar\n_casado/sdn- abstrac\ons The Future of Networking, and the Past of Protocols

Many bullet points on next few slides are from this talk
46

The Shenker View: the gist

Network control planes need abstrac\ons
Abstrac\ons solve architectural problems and enable evolvability Todays layers (L2, L3, ..) are good abstrac\ons but terrible for control interfaces

Networks work because we can master complexity

but what we should be doing is extrac\ng simplicity, with the right abstrac\ons
47

Programming Made the Transi\on

Machine Languages: no abstrac\ons Higher-level languages, OS + other abstrac\on
les, virtual memory, data structures,

Modern languages: even more abstrac\ons

objects, garbage collec\on, threads, locks,

Abstrac\ons simplify programming: they make it easier to write, maintain, and reason about programs. Could networking follow this same path?

48

Forwarding Abstrac\on
Forwarding behavior specied by a control program. Possibili\es: x86, MPLS, OpenFlow

49

State Distribu\on Abstrac\on

Control program should not have to handle distributed-state details Proposed abstrac\on: global network view Control program operates on network view
Input: global network view (graph) Output: congura\on of each network device

Network OS provides network view

Short version: programs operate on graphs

50

Specica\on Abstrac\on
Give control program abstract view of network Provide enough detail to specify goals, but not to implement them

51

What is SDN, opt 3

My view:

Opening Up Design Axes

52

Dene SDN not by what it looks like or how we think about it, but the exibility it provides.
53

SDN opens up implementa\on design axes.

54

An SDN is any network that gives us the exibility to choose between points on the following design axes.
55

Centralized vs Distributed Control

Centralized Control
Controller

Distributed Control
Controller

OpenFlow Switch

OpenFlow Switch

Controller

OpenFlow Switch

OpenFlow Switch

Controller

OpenFlow Switch

OpenFlow Switch

56

Microow vs. Aggregated

Microow

Every ow is individually set up by controller Exact-match ow entries Flow table contains one entry per ow Good for ne grain control, policy, and monitoring, e.g. campus

Aggregated

One ow entry covers large groups of ows Wildcard ow entries Flow table contains one entry per category of ows Good for large number of ows, e.g. backbone

57

Reac\ve vs. Proac\ve (pre-populated)

Reac\ve

First packet of ow triggers controller to insert ow entries Ecient use of ow table Every ow incurs small addi\onal ow setup \me If control connec\on lost, switch has limited u\lity Extremely simple fault recovery

Proac\ve

Controller pre-populates ow table in switch Zero addi\onal ow setup \me Loss of control connec\on does not disrupt trac Essen\ally requires aggregated (wildcard) rules

58

Virtual vs Physical
Virtual

Assumes congurable switching within a host: in the OS or hypervisor Soyware! Memory, processing, arbitrary modica\ons Massive ow rates Limited to the hardware below

Physical

No assump\on of soyware changes; unmodied end hosts Greater control over expensive forwarding resources

59

Fully Consistent vs Eventually Consistent

Fully Consistent

Certainty about state Consistent state is harder to scale Easier to reason about state and its transi\ons May eliminate route aps

Eventually Consistent

Uncertainty about state now, but eventually converges Probabilis\c state is easier to scale Introduces the possibility of long-lived route aps and unstable control systems

60

Heres a picture

61

These axes, today (BGP):

Centralized Microflow Reactive Virtual Fully Consistent Distributed Aggregated Proactive Physical Eventually Consistent

Many choices arise from assuming decentralized administration

62

These axes, for Ethane

Centralized Microflow Reactive Virtual Fully Consistent Distributed Aggregated Proactive Physical Eventually Consistent

Many choices arise from assuming centralized administration

63

These axes, for Hedera:

Centralized Microflow Reactive Virtual Fully Consistent Distributed Aggregated Proactive Physical Eventually Consistent

Hedera controls a data center network. See NSDI 10 paper for more details.
64

Your SDN network or applica\on?

Centralized Microflow Reactive Virtual Fully Consistent Distributed Aggregated Proactive Physical Eventually Consistent

Hybrid approaches often have desirable properties; can use these now.
65

Example applica\ons and their places on the scales

66

High-order bit of SDN: adds exibility to control-plane implementa\on choices

67

SDN examples using OpenFlow

Stanford Demos Wireless mobility VM mobility/migra\on Network virtualiza\on Power management Hardware Load balancing Trac Engineering Others

Many applica\ons benet from exibility

Removing spanning tree Network visualiza\on Network debugging Packet-circuit convergence Home networks Flexible access control Scale-out routers Scale-out data centers

68

What SDN really means is up in the air. Heres a good deni\on, though: Soyware Dened Networking (SDN) is a refactoring of the rela\onship between network devices and the soyware that controls them.
[Paraphrased from the HotSDN 12 Solicita\on]

69

The SDN Stack

70

The SDN Stack

oyrace oops openseer

Monitoring/ debugging tools

ENVI (GUI)

LAVI

n-Cas\ng Trema Maestro

Applica\ons Controller Slicing Soyware

Broadcom Ref. Switch

NOX

Beacon FlowVisor Console

FlowVisor

Commercial Switches
HP, NEC, Pronto, Juniper.. and many more

Soyware Ref. Switch OpenWRT

NetFPGA

PCEngine Open vSwitch WiFi AP

OpenFlow Switches
71

Switches (all OpenFlow 1.0 for now)

72

Current SDN hardware (as of ~2010)

Juniper MX-series NEC IP8800 WiMax (NEC)

HP Procurve 5400

Netgear 7324

PC Engines

Pronto 3240/3290

Ciena Coredirector

Ask your vendors

73

Commercial Switches
Vendor HP ProCurve Models Virtualize? 5400zl, 1 OF instance 6600, + per VLAN Notes -LACP, VLAN and STP processing before OpenFlow -Wildcard rules or non-IP pkts processed in s/w -Header rewri\ng in s/w -CPU protects mgmt during loop -OpenFlow takes precedence -Most ac\ons processed in hardware -MAC header rewri\ng in h/w -No legacy protocols (like VLAN and STP) -Most ac\ons processed in hardware -MAC header rewri\ng in h/w Pic

NEC

IP8800, 1 OF instance + per VLAN 3290, 1 OF instance 3780, per switch 3920, +

Pronto/ Pica8

74

Open Switches
Name OpenFlow Reference Open vSwitch Indigo Lang C PlaQorm(s) License Linux OpenFlow License Original Author Stanford/ Nicira Ben Pfa/ Nicira Dan Talayco/ BigSwitch Pica8 Notes not designed for extensibility

C/ Python C/Lua

Linux/BSD? Apache 2.0

In Linux kernel 3.3+

Linux-based GPL v2 Hardware Switches Linux-based ? Hardware Switches

Bare OpenFlow switch

Xorplus

Support

75

Ignore these slides. Ask your vendor for the latest.

76

What you cannot do with OpenFlow v1.0

Non-ow-based (per-packet) networking
ex. Per-packet next-hop selec\on (in wireless mesh) yes, this is a fundamental limita\on BUT OpenFlow can provide the plumbing to connect these systems

Use all tables on switch chips

yes, a major limita\on (cross-product issue) BUT OF version 1.1 exposes these, providing a way around the cross-product state explosion

77

What can cannot do with OpenFlow v1.0 (2)

New forwarding primi\ves
BUT provides a nice way to integrate them through extensions

New packet formats/eld deni\ons

BUT extensible matches are coming in v1.2/1.3+

Op\cal Circuits
BUT eorts underway to apply OpenFlow model to circuits

Low-setup-\me individual ows

BUT can push down ows proac\vely to avoid delays

Where its going

OF v1.0: released end of 2009: Into the Campus OF v1.1: released March 1 2011: Into the WAN
mul\ple tables: leverage addi\onal tables tags and tunnels: MPLS, VLAN, virtual ports mul\path forwarding: ECMP, groups

OF v1.2: approved Dec 8 2011: Extensible Protocol

extensible match extensible ac\ons IPv6 mul\ple controllers
79

Controllers

80

Open Controllers
Name OpenFlow Reference NOX Lang C PlaQorm(s) License Linux OpenFlow License GPL Original Author Stanford/ Nicira Nicira Notes not designed for extensibility

Python, Linux C++ Java Win, Mac, Linux, Android Win, Mac, Linux Linux

ac\vely developed

Beacon

GPL (core), David FOSS Licenses Erickson for your code (Stanford) LGPL Zheng Cai (Rice) NEC

run\me modular, web UI framework, regression test framework

Maestro

Java

Trema

Ruby, C

GPL

includes emulator, regression test framework virtual IP rou\ng as a service

81

RouteFlow

Linux

Apache

CPqD (Brazil)

Open Controllers (2)

Name Lang PlaQorm(s) License Original Author Notes Library OpenFaucet Python Mirage POX Floodlight OCaml Python Java Any Any BigSwitch, based on Beacon

Too many to easily list of keep track of

82

Growing list.
Mar\n Casado recently added a list: h]p://yuba.stanford.edu/~casado/of-sw.html

83

Related Research
DIFANE
Rule par\\oning for controller-less ow inser\on

UCSD Fat Tree Series: Scalable Commodity Data Center, PortLand, Hedera
Scale-out data centers that use OpenFlow

Tesseract
Centralized WAN in the 4D Architecture

ONIX
Fault-tolerant controller plaGorm from Nicira, Google, NEC

DevoFlow
Prac\cal scalability limits to OpenFlow and modica\ons 84 to get around them

Related Research
Frene\c/Ne]le
Func\onal Reac\ve Programming for more composable, reusable controller code

Resonance
State-machine-based network control

Consistency Primi\ves
Per-packet or per-ow rou\ng guarantees to simplify network versioning

85

Up Next:
Come back at 10:45 for: The SDN Stack Part 2: Virtualiza\on and SDN Applica\ons

86

Ge#ng Started
(1) Copy to your hard disk from a USB Key or DVD:
Copy needed les (VirtualBox, terminal, possibly an X server) for your plaGorm (Win/Mac/Linux) Copy Java 6 and Eclipse for your plaGorm, if you want to use Java Copy VM image: OpenFlowTutorial-101311.zip Pass on the DVD or USB key to someone else!

(2) Unzip OpenFlowTutorial-101311.zip

(3) Point browser to instruc\ons:

h]p://www.openow.org/wk/index.php/OpenFlow_Tutorial (note the underscore)

You should NOT need to download any large les spare the WiFi!
87

Virtualiza\on and SDN Applica\ons

88

Virtualiza\on Rob Sherwood, (BigSwitch)

89

Current Trials
68 trials/deployments spanning 13 countries

90

Internet2 OpenFlow deployment ini\a\ve. 35+ 100G POPs, na\onwide.

Applica\ons of SDN

92

OpenFlow Demonstra\on Overview

Topic
Network Virtualization Hardware Prototyping Load Balancing Energy Savings Mobility Traffic Engineering Wireless Video

Demo
FlowVisor OpenPipes PlugNServe ElasticTree MobileVMs Aggregation OpenRoads
93

Demo Infrastructure with Slicing

WiMax WiFi APs

OpenFlow switches

Flows

Packet processors

94

FlowVisor Creates Virtual Networks

OpenPipes Demo OpenRoads Demo PlugNServe Load-balancer

Each demo presented here runs in an isolated slice of Stanfords produc\on network.
OpenFlow Switch

OpenFlow Protocol

OpenFlow Protocol

FlowVisor

OpenPipes Policy

OpenFlow Switch

OpenFlow Switch

FlowVisor slices OpenFlow networks, crea\ng mul\ple isolated and programmable logical networks on the same physical topology.
95

OpenPipes
Partition hardware designs

Plumbing with OpenFlow to build hardware systems

Mix resources

Test

96

Load-Balancing Web Trac using OpenFlow Goal: Load-balancing requests in unstructured networks
What we are showing
OpenFlow-based distributed load-balancer
Smart load-balancing based on network and server load Allows incremental deployment of addi\onal resources

Plug-n-Serve:

OpenFlow means
Complete control over trac within the
network Visibility into network condi\ons Ability to use exis\ng commodity hardware

This demo runs on top of the FlowVisor, sharing the same physical network with other experiments and produc.on trac.

97

Dynamic Flow Aggrega\on on an OpenFlow Network

Scope
Dierent Networks want dierent ow granularity (ISP, Backbone,) Switch resources are limited (ow entries, memory) Network management is hard Current Solu\ons : MPLS, IP aggrega\on

How OpenFlow Helps?

Dynamically dene ow granularity by wildcarding arbitrary header elds Granularity is on the switch ow entries, no packet rewrite or encapsula\on Create meaningful bundles and manage them using your own soyware (reroute, monitor) Higher Flexibility, Be[er Control, Easier Management, ExperimentaEon

98

Elas\cTree:

Reducing Energy in Data Center Networks

Shuts o links and switches to reduce data center power Choice of op\mizers to balance power, fault tolerance, and BW OpenFlow provides network routes and port sta\s\cs The demo: Hardware-based 16-node Fat Tree Your choice of trac pa]ern, bandwidth, op\miza\on strategy Graph shows live power and latency varia\on
demo credits: Brandon Heller, Srini Seetharaman, Yiannis Yiakoumis, David Underhill
99

100

Intercon\nental VM Migra\on

Moved a VM from Stanford to Japan without changing its IP. VM hosted a video game server with ac\ve network connec\ons.
101

openow.org/videos

102

Play Videos: Wireless Handover Load balancing

103

Times are changing

No longer what can I now do that I couldnt do before Clouds Cost eciency Manageability Programmability

104

Where SDN make sense (and doesnt)

[from Martin Casados blog: http://networkheresy.wordpress.com/2011/11/17/isopenflowsdn-good-at-forwarding/]

So the ques\on really boils down to whether the algorithm needed to compute the datapath state is easily distributed ... if the state management algorithm has any of the following proper\es it probably isnt, and is therefore a poten\al candidate for SDN. It is not amenable to being split up into many smaller pieces (as opposed to fewer, larger instances). The limi\ng property in these cases is oyen excessive communica\on overhead between the control nodes. It is not amenable to running on heterogeneous compute environments. For example those with varying processor speeds and available memory.
105

Where SDN make sense (and doesnt)

[from Martin Casados blog: http://networkheresy.wordpress.com/2011/11/17/isopenflowsdn-good-at-forwarding/]

It is not amenable to rela\vely long RTTs for communica\on between distributed instances. In the purely distributed case, the upper bound for communica\ng between any two nodes scales linearly with the longest loop free path. The algorithm requires sophis\cated distributed coordina\on between instances (for example distributed locking or leader elec\on)

106

SDN stack
Monitoring/Debugging Tools Applica\ons Controllers Virtualiza\on Switches Topology

107

Up Next:
1:00 - 3:00pm Controllers: NOX/POX (Murphy) Trema (Hideyuki) Beacon (David) Floodlight (Rob) RouteFlow (video) + more Q&A
108

Q & A
109

Ge#ng Started
(1) Copy to your hard disk from a USB Key or DVD:
Copy needed les (VirtualBox, terminal, possibly an X server) for your plaGorm (Win/Mac/Linux) Copy Java 6 and Eclipse for your plaGorm, if you want to use Java Copy VM image: OpenFlowTutorial-101311.zip Pass on the DVD or USB key to someone else!

(2) Unzip OpenFlowTutorial-101311.zip

(3) Point browser to instruc\ons:

h]p://www.openow.org/wk/index.php/OpenFlow_Tutorial (note the underscore)

You should NOT need to download any large les spare the WiFi!
110

Break
Come back at 1:00 for Controllers

111

Controllers
112

POX: Murphy McCauley Beacon: David Erickson Trema: Hideyuki Shimonishi Floodlight: Rob Sherwood
113

VM Setup/ Overview
114

[Hands-on Tutorial] Overview

115

TutorialFlow

116

Stu youll use

Ref Controller / NOX / Beacon / Trema / Floodlight Open vSwitch Mininet iperf tcpdump Wireshark

117

Tutorial Setup
Controller c0 port6633 OpenFlow Tutorial 3hosts-1switch topology
loopback (127.0.0.1:6633)

s1 OpenFlow Switch
s1-eth0 s1-eth1

loopback (127.0.0.1:6634)

dpctl (user space process)

s1-eth2

h1-eth0

h3-eth0

h4-eth0

10.0.0.2

h2

10.0.0.3

h3

10.0.0.4

h4

virtual hosts

118

Pick a controller!

119

POX: Murphy McCauley Beacon: David Erickson Trema: Hideyuki Shimonishi Floodlight: Rob Sherwood
120

Hands-on Tutorial
Instruc\ons at:
www.openow.org/wk/index.php/OpenFlow_Tutorial

121

Q&A
122

Deployment Experiences
123

Deployment: Srini Seetharaman

124

Wrap-Up
125

OpenFlow/SDN Tutorial A[endees

180 160 140 120 100 80 60 40 20 0

1/1/08 1/1/09 1/1/10 1/1/11 1/1/12

A]endees

Expon.(A]endees)
126

Growing Community
Vendors and start-ups Providers and business-unit

More...
Note: Level of interest varies

More...
127

Closing Thoughts
SDN & OpenFlow Open Source Ecosystem SDN Switches and Controllers are available, used, and improving: open & closed These are the early stages for SDN

128

Get involved!
Ask and answer ques\ons on mailing lists:
openow-discuss openvswitch-{discuss/dev} ONF forums

Share via wiki, github, etc. Submit bug-reports and/or patches to OF reference implementa\on and Open vSwitch Release open-source applica\ons Write a new controller!
129

This tutorial wouldnt be possible without:

Speakers
Rob Sherwood David Erickson HIDEyuki Shimonishi Murphy McCauley Srini Seetharaman

130

ACKs
(acknowledgements)
131

This tutorial wouldnt be possible without:

OpenFlow Experts
Tatsuya Yabe Nikhil Handigol TY Huang James Hongyi Zeng Bob Lantz Masahiko Takahashi Ali Al-Shabibi Ali Yahya Glen Gibb
132

This tutorial wouldnt be possible without:

Past slides from:
Nick McKeown Rob Sherwood Guru Parulkar Srini Seetharaman Yiannis Yiakoumis Guido Appenzeller Masa Kobayashi Sco] Shenker others
133

This tutorial wouldnt be possible without:

Behind-the-scenes organizers:
Flora Freitas Asena Gencel Guru Parulkar

134

SDN Team at Stanford