About us: Management

Computer Networks
CS 552 Professor: Badri Nath
http://www.cs.rutgers.edu/~badri
badri@cs.rutgers.edu
Badri Nath Office Hours: Wednesday 1:30-3:30 PM; CORE 320
TA: Long T Le
Rutgers University
longtle@cs.rutgers.edu
badri@cs.rutgers.edu Office Hours: Tuesday 10:00-11:00 AM; Hill 202
SPRING 2017 Course info
http://www.cs.rutgers.edu/~badri/552.html

1 2

Course Web Page Sakai Web page

https://sakai
Course schedule Will submit reviews or answer to mini-quizzes
Reading list (based on the paper) online
Lecture notes Course announcements
Announcements Written Homeworks
Assignments
Project ideas
Exams

3 4

1
Course Goals
Course Materials

Understand the basic principles of computer networks, in Research papers

particular the Internet
Study new concepts, design principles in network Links to pdf on Web page
protocols and design Combination of classic and recent work
How to do network research ~30 papers
How to determine what is important
What are the trends Optional readings
Internet of things (IoT), Datacenter, cloud, SDN, Recommended textbooks
Cloud vs edge:
connected devices/home, connected vehicles For students not familiar with networking
If software is eating the world, networking is enabling it
Peterson & Davie (4th edition)
What are the economics, technology that is driving innovation
Cost, performance, energy, availability, security Alternative: Kurose & Ross
Sharing Economy: airbnb, uber
5 6

Reading papers Books

Understand /identify the basic idea Computer Networks: A Systems Approach, 4th Ed. (2007), by
Larry Peterson and Bruce Davie.
What is the problem that the paper tackles?
Computer Networking: A Top-Down Approach Featuring the
What kind of a paper? Internet, 5th Ed. (2010), by James F. Kurose and Keith W. Ross.
Performance, vision, new direction/protocol paper TCP/IP Illustrated, Volume 1: The Protocols by W. Richard
Summarize key idea Stevens.
Unix Network Programming: Networking APIs: Sockets and
+ve aspects of the paper XTI (Volume 1) by W. Richard Stevens.
New, breakthrough, incremental, Advanced Programming in the Unix Environment by W.
-ve aspects of the paper Richard Stevens, Addison-Wesley, 1993.
Readability, Assumptions (valid?), scaling issues (does it I or 2 recommended
scale), implementation (has it been implemented),
measurements (problems?)
7 8

2
Grading
Class Coverage

15% Paper summaries/reviews/HWs based on Papers Quick overview of undergraduate networking

A subset of the papers will be assigned for submitting summary/critique Pre requisite: 352 or equivalent
All papers assigned should be read as quizzes/Hws will be based on these
papers
Students expected to know
Link layer, basic IP routing, TCP,
35% Programming project (two-person)
Focus on Advanced topics in networking (from
20% Mid term papers in recent SIGCOMM, NSDI)
30% Final Course will deal with:
Honor code Services and Protocols
All submitted work should be yours Investigate protocol trade-offs, new cost models
You are all grad students!! New Workloads, new technologies, new services
Scheduling, QoS, Load balancing
Software defined networking
9 10

Class etiquette Lecture Topics

Traditional Recent Topics

Cell phones in off position Layering Energy, $$
No FB status updates in class Internet architecture Considerations
If you need to surf while in class (I prefer not), Routing (IP, BGP) Measurements
do not disturb your neighbors Transport (TCP)
CDN/Video/DASH
Stop me anytime to ask questions Datacenter networking
Prof may not know the answer!!
Cloud(IaaS)
This is a graduate class, student participation
Software defined
in class is important networking
Challenge the class, the prof, and ideas in papers
11
Trends: Desktop/Internet Mobile/Cloud Edge computing 12

3
What is a Network? Why Networks?

Carrier of information between 2 or more entities Availability of Resources

Some carry objects/people (postal, air, surface transport)
Resources become available regardless of the users
Most important is the services offered
physical location (server based, peer2peer)
User expectation of service
Latency, cost, reliability, service interface, others Load Sharing/utilization
We focus on computer networks Jobs processed on least crowded machine
Interconnection may be any medium capable of communicating information:
Resources can be shared
copper wire
Lasers (optic fibre) High Reliability
Microwave
Cable (coax) Alternative source of supply (multiple copies)
wireless Human-to-Human Communication
satellite link
Example: Ethernet, Wifi, 3G, LTE e.g., on-line world, e-commerce

13 14

Domain name growth What is Internet Technology?

What is an internet?
Network of networks
What is the Internet?
A global internet based on the IP protocol
To what does Internet technology refer?
Architecture, services, interfaces, and
protocols

15 16

4
 Impact on society

20 M msgs/ minute
2.4 M Queries/ minute
Scale Scale Scale
Net neutrality
ISPs should not discriminate
For and against case
Nations laws and Internet
Regulation across national boundaries
Privacy
Ears and Eyes in your house connected to the net.
Who can ask for that?
Is Alexas (Amazon) recordings available to others?
Content creation , ownership, distribution, online piracy
Cyber Warfare N Korea, Hacking (political objectives)

17 18

Internet Players Internet Growth

Users, people who use the applications

Everyone (mom and pop, kids)
get something done (hopefully useful)
Designers
You: protocol design and implementation
Scale, performance, cost, incremental deployment
Service Providers
Administrators and ISPs
Datacenter operators
Provider-customer versus peer-to-peer
Management, revenue, deployment
Market/business models for the Internet
Consumer to consumer (ebay, match.com, craigslist, airbnb), Business
to consumer (amazon, orbitz, google, netflix, hulu), Business to
business (getty, harvest, google), Consumer to business ( hot jobs,
monster,linkedin)

19 20
Mary meeker : Internet trends May 2015

5
 Internet Growth WWW growth

21 22

Souce: Internet trends May 2013, Mary Meeker: Astounding growth of Google

23 24

6
Facebook growth Social Networking/Messaging

25 26
MAU: Number of unique users in the past 30 days

Messaging growth Initially: Only kids!!

27 28

7
Now: everyone; grandpa, grandma Mobile Phone usage

29 30

Mobile OS trends

31 32

8
Communication trends Video Content Growth

Can nodes be selected dynamically? Mary Meeker Internet trends 2013

33 34

Data Producers everywhere

35 36

9
 Reimagining shared economy Whats hot?

Renting a place Software Defined

Hailing a cab Renting your car Networking (SDN)

Embedded
Mobile &Wireless Internet of things (IoT)
Internet Bodily of Things (IoB)

Internet

Social Networking Bots

Data centers
Edge vs Cloud
http://a16z.com/2016/12/16/the-end-of-cloud-computing/

On demand food delivery Supply/Demand Exchanges 37 38

Circle of networking Edge is where the action will be

Architecture
Car: data center on wheels
Drones: data center on wings
Desktop Robots: data center with arms and legs
mainframes
Flying cars?
Edge computing
Mobile/cloud Mainframe to desktop to cloud to edge
Sense , infer, query, act
Watch the Video By Levine: return to the edge
http://a16z.com/2016/12/16/the-end-of-cloud-computing/
Action is all at the edge: cars, drones, robots
39 40

10
Device Growth: connected Smart cities/Smart Campuses
Software defined urban platforms
If software is eating the world, networking is enabling it

41 42

Objective of networking How to communicate?

End-hosts to communicate Circuit switching

Applications running on end-hosts Establish a connection before communicating
Different technologies POTS (plain old telephone system)
Dedicated pipe for the duration of the session
Different protocols
Packet switching
Different Services Multiplex communication from different sources
Every packet is self contained
Efficient use of resources
NO guarantees on performance

43 44

11
How to handle different How to locate a node?
networks?

Many differences between networks Naming, discovery and routing

How to translate between various network Network elements needed to support directory
technologies? Network elements needed to support forwarding
towards destination
Have a common protocol for inter network
Scalable
communication
Reliable
IP
A set of rules with a well-defined interface

45 46

How to meet application

demands? Lots of Functions Needed

Corruption? Link
Need error detection and correction Multiplexing
Reliability Routing
Data lost? Addressing/naming (locating peers)
Overload Reliability
Congestion control
Flow control
Security
Fragmentation
Encryption, authentication
Etc.
47 48

12
ISO OSI Layering Architecture Problems

Host A Host B
Application Application Protocol Application
Seven layers not widely accepted
Layer Layer
Standardized before implemented
Presentation Presentation Protocol Presentation
Layer Layer
Top three layers fuzzy
Session Session Protocol Session
Layer Layer Internet or TCP/IP layering widespread
Transport Transport Protocol Transport
Layer Layer

Network Network Network Network

Layer Layer Layer Layer

Data Link Data Link Data Link Data Link

Layer Layer Layer Layer

Physical Physical Physical Physical

Layer Layer Layer Layer 49 50
Router Router

TCP/IP Layering Architecture Internet design philosophy

A simplified model #1 Functionality at the edge as opposed to core

The network layer In Telephone network it is the opposite
Application
Hosts drop packets into
this layer, layer routes
Any new service, the phone company has to provide
Transport (TCP/UDP) towards destination- only Edge device is dumb
promise- try my best Smart device at the edge means programmability
Internet/Network The transport layer
(IP) New services can be supported, drives innovation
Reliable/unreliable byte-
oriented stream VOIP (SIP), IM
Host-to-Net
Cathedral vs Bazaar

51 David Clark, The design philosophy of the DARPA internet protocols, 1998 52

13
Implications (cathedral vs # 2 Multiplexed utilization
bazaar)

Bazaar Best effort, packet switching

Keep the network simple
Edge is programmable Packets may be lost, corrupted, out-of-order
Let the end host implement any other requirements
Nimble, novel applications Want reliability?
Cathedral Retransmit from sender
Packets self contained
Core elements still rigid Can take different routes
Standards, slow evolution Different transfers on the same link
Stateless in the core
Cant do anything radical End hosts can maintain state
Fate-sharing (If I Die, my state will die but will not affect
Where is programmability in the system? others)

53 54

#3 Support multiple networks TCP/IP Layering Architecture

IP over anything, anything over IP

Run over any type of link
Build any end-to-end protocol over IP Host A Host B
http
Application Protocol http
Application
Transport Protocol (TCP)
TCP Ethernet

UDP TCP Transport

Network
IP Network
IP Network
IP Network
Layer Layer Layer Layer

IP Host-to- Host-to- Host-to- Host-to-

Net Layer Net Layer Net Layer Net Layer
Ethernet OC3 Ethernet
Network Interface

55 56

14
#4 distributed management

Need only know local information

Information distributed over different nodes
Scalable
No single hot spot
Distributed functionality-- Roles
Different entities manage different parts of the
system
Impact on naming, routing, addressing
Hierarchical topology but this too is changing
Local and Global management authorities
57 58

Facebook DC
https://code.facebook.com/posts/360346274145943/introducing-data-center-
fabric-the-next-generation-facebook-data-center-network

Major elements:
CDN Content distribution networks
IXPs Internet Exchange points
Datacenters

59 60
4x40 G RSW

15
Internet Design Principles End-to-end argument

Scale Saltzer, reed and clark [1984]

End-to-end arguments in system design
Protocols should work in networks of all sizes and
Main idea
distances
If a function can only be completely and correctly implemented with the
Incremental deployment knowledge and help of the applications standing at the communication
end points. Hence providing this function in the subsystem is not
New protocols need to be deployed gradually possible
Heterogeneity Complexity at the edges as opposed to the core
Simply stated, the argument suggests that functions placed at
Different technologies, autonomous organizations the low levels of a system may be redundant or of little value
when compared with the cost of providing them at that low
End-to-end argument
level.
Networking functions should be delegated to the edges; Dont force feature, service, restriction on the end points
application knows best

61 62

Communication system End-to-end argument

An end system connected by a Functions placed at lower level implies

communication subsystem specific problems being solved in a general
way
Questions?
Best aim:
Who is responsible for a given function
Simple lower layer with smart end points
Subsystem?
Basic and general functions at the lower layers
End units?
Gives flexibility
Or both (redundant) or jointly?

63 64

16
e2e argument e2e tradeoffs

Low level system may not Performance New business models

have all the information to
implement the given
functionality
Implement only for
performance (wireless links)
Low level system shared by
all applications what if the
application does not need
the feature
Duplicated effort
Network caching, redirection, proxy transcoding
Should not impact Wireless Application protocols
applications that do not use
that functionality Gateway provides a box for content translation
Network redirection
Network level switch for load balancing
Balance between performance, layering, e2e
argument

65 66

New metrics Typical Datacenter networking

Energy/power
Always-on system consumes a lot of power

System Performance- Latency, throughput, Availability

How to guarantee performance on shared Infrastructure? IaaS
Latency, throughput, availability Request=<L,B, N>
How to measure?
Cost <$>
Hardware cost is falling, software lease cost
Opex vs capex debate
Cloud/Software as a Service (SaaS) models
Cost proportional to usage
How to bid? How to optimize?

67 68

The cost of a cloud: Research Problems in data center networks by Albert Greenberg et.al, CCR

17
Data center traffic (Of Mice and DC network evolution
Elephants): Latency, BW, $ as metric

50% of flows < 100

5% of flows >10 MB
KB
30% of bytes
<5% of bytes

>10MB

<100KB

Srikanth Kandula- Nature of Datacenter traffic, IMC2009 69 70

Mohammad Alizadeh, et.al., -Less is more NSDI 2012,

Datacenter cost Facts and Figures

[Quereshi09]
Servers are power hungry (annual electricity bills)
Servers Power Infrastructure Network
45% 15% 25% 15%

50,000 server @ 3K a pop, 5% cost of money, 3YR

52.5 M/Yr cost
Power cost
Power to run the IT equipment
Power to run cooling, UPS etc Overhead
PUE=Total power /IT power
1.2 ideal -- 20% overhead
Typically 2 to 3 PUE Air conditioning costs enormous

71 72

18
Energy-proportional metric
Energy Proportional Computing

Can we design networks that consume power

proportional to utilization?
The Case for Energy-Proportional Computing,
Luiz Andr Barroso, Urs Hlzle, IEEE Computer,
vol. 40 (2007).
The Case for
Energy-
Proportional
Computing,
Luiz Andr
Barroso,
Urs Hlzle,
IEEE Computer
December 2007

Figure 1. Average CPU utilization of more than 5,000 servers during a six-month period. Servers
73 are rarely completely idle and seldom operate near their maximum utilization, instead operating 74
most of the time at between 10 and 50 percent of their maximum

Throughput proportional
Energy Proportional Computing
fabric

util power EE
0.15 0.6 .25
0.4 0.7 .57 Doing nothing well?
Still power

Does throughput offered rise/fall in

proportional to traffic sources/sinks? Energy Efficiency =
Utilization/Power
Fat-free topologies paper in Hotnets 2016
Figure 2. Server power usage and energy efficiency at varying utilization levels, from idle to
peak performance. Even an energy-efficient server still consumes about half its full power
75
when doing virtually no work. 76

19
 Energy/ Power metric
Energy Proportional Computing
If Cost/Green Conscious
Can we do this for Make network elements less power hungry
Networking [1]Gupta & Singh Greening of the Internet SIGCOMM
Infrastructure? Doing nothing Design for 2002
VERY well wide dynamic [2] Energy Proportionality of an enterprise network, Priya
power range and Mahadevan, et.al, Green Networking August 2010
active low power
modes Take advantage of lower power rates elsewhere,
time-of-day
Energy Efficiency = [3] A. Qureshi, R. Weber, H. Balakrishnan, J. Guttag, B.
Utilization/Power
Maggs, "Cutting the Electric Bill for Internet-Scale
Systems" SIGCOMM 2009
Figure 4. Power usage and energy efficiency in a more energy-proportional server. This
server has a power efficiency of more than 80 percent of its peak value for utilizations of
30 percent and above, with efficiency remaining above 50 percent for utilization levels as
77 78
low as 10 percent.

What is the energy consumption of Facts and Figures [Gupta&Singh-03]

the internet ?

Energy consumed by networking Device Approximate Number Total

equipment such as routers, switches, Deployed AEC TW-h
hubs etc Hubs 93.5 Million 1.6 TW-h
Does not include hosts LAN 95,000 3.2 TW-h
Internet energy consumption controversial Switch
data WAN 50,000 0.15TW-h
Switch
Router 3,257 1.1 TW-h

79 80
P=E/T 1W = 1 J/S 1 Kw-H = 3600000 joules =10 x100W bulbs for 1 Hr

20
More Numbers Enterprise networks

Total energy consumed by networking devices Network Switches hubs

annually in 2000 (US): 6.05 TW-h routers under utilized
Amounts to about 0.07 % of total U.S. energy Many units not energy
expenditure proportional
Expected increase: +1 TW-h by 2005 Turn off ports in
proportion to b/w
Note: This does not include energy consumed
demands
by hosts, UPS supplies or cooling equipment.
Timescales?

81 82

So, why bother saving energy?

How to save energy ?

3 reasons:
Low-Energy Hardware Design:
1. Current energy inefficiencies Use hardware components with low power modes of
a) Wired vs. wireless energy costs operation
b) 6 TW-h ~ 1 nuclear reactor Lower the clock frequency of the components, use
c) Extrapolate to World ~140 nuclear reactors DVS and other methods during low demand
2. Enable greater deployment Energy-Aware APIs give control to software
a) Similar connectivity in India would require 4.75% of total Architecture that allows selective powering off
energy budget
3. Enable longer operation times during events of disaster Energy-Aware Protocol Design:
a) Recent Grid failure in NE US/Canada Node-level algorithms for sleeping
b) Frequent power outages in most of the world Route aggregation and other global techniques to
inform devices when and for how long to sleep

83 84

21
Where to save energy in a device ?
Key Questions

For how long can components sleep?

Memory state transition times, energy spike
Main CPU How is the decision to sleep taken ?
traffic activity level, in isolation (uncoordinated), global (coordinated),
Switch fabric or bus backplane edge or backbone device, transit or stub network
How to distinguish sleep vs. failure ?
Line cards (designs ranging from simple to should not trigger network reconfiguration in sleep state vs. failure state
complex with ASICs or network processors to How to wake up a device ?
process packets) at fixed intervals, on packet arrival, account for protocol timers
Impact on protocol behavior?
Bang for the buck long sleep times, slower propagation of topology changes

85 86

More on sleeping. Computation Placement

Uncoordinated sleeping:
Decision to sleep based on individual traffic levels alone Energy cost varies by location
Inform nearest neighbors only
Sleep time limited by protocol hello message timer
Energy cost varies by hour of the dat
May trigger network reconfiguration in case of missed protocol
messages Can we push computation to a geographically
Coordinated sleeping: distant place to save energy?
Decision taken on a network-wide basis
Need algorithm to pre-compute the optimal sleep time, but Can we exploit time difference?
computation costs increase
Hello message frequency can be adjusted, may take longer to detect Peak vs non-peak power rates
changes in network topology
Sleep longer intervals, but forwarding tables may be outdated
Reroute all traffic through one route, shutdown other routes
Introduces delay and packet loss in case of sudden traffic burst

87 88

22
Exploit spatial-temporal price Research
variation

Prices vary from place to place Modification of protocols at layers 2 and 3 to

Coal vs nuclear incorporate sleep modes
Price varies with time of day Study impact of modifications on end-to-end
delay and performance
Peak vs off-peak
Develop energy models for routers and
switches
Study the algorithmic problems of
how/when/how long to sleep
Load migration: tolerable latency vs cost
savings
89 90

The tail at scale Variability in latency

Curse of the long tail
99% of the requests finish < 10 msec
1% of the request > 1 sec
A large fanout, more requests will need > 1 sec Shared resources
Touch 100 servers: Prob not one of them is cursed =0.99100 =0.37
63% of requests take > 1 sec Background jobs
Queuing
Maintenance jobs (checkpoint)
R
e
Large Fanout: Multiplier effect
q
u
e
s
t

Latency The tail at Scale by Dean and Barroso, CACM, Feb 2013, Vol. 56, NO.2

91 92

23
 Use redundancy Hedged requests
S1 S2

Redundancy effectively used to improve

r5
availability (RAID, replicated servers) r5
Now use to reduce variability in latency
Response time important in many services
Word completion, spellchecking, document selection client
Idea: Send requests to multiple servers
Pro: Hit one server with low delay Send requests to multiple servers
Con: more resources consumed due to duplicate Cancel the other after first response
work (increase in queueing delay) Techniques to reduce overhead:
Challenge: reduce latency without undue Send second after some delay (d>95 percentile)
overhead Send second at a lower priority
93 94

Tied requests Results

S1 S2
cancel

r5
Hedged requests: 100 servers
r5 Send hedged request after 10 msec delay
Reduced 99.9 percentile latency from 1800 msec
<r5,S1>
to 74 msec
<r5,S2>
Sends only 2% more requests
Send requests to multiple servers (with id of replicas) Tied requests
Server cancels other requests when the request scheduled
40% reduction at the 99.9 percentile latency
Still both requests may be served (queue empty)
Techniques to reduce overhead: Key takeaway: Predictability from unpredictable parts
Send second after some delay (d> RTT of message in Network) Can we do this for network security?

95 96

24
Computing cost declining: own or rent Cloud services

Amazon Web services

Google cloud
Microsoft Azure

Infrastructure as a service Platform as a service

97 98

Rutgers pays about 30K a month for ISPs

Amazon cloud pricing Cost of cloud

On-demand: pay by the hour, no upfront
payment (0.023 per hour to 0.094 per hour)
# CPUs, memory size, disk space etc varies
Spot pricing
Bid for spare capacity, bid price < spot price,
process is preempted
Reserved Pricing
1 yr to 3 yr commitment; upfront payment
Large instance $549 plus 0.063 per hour
99
Many cost models for compute resources
Storage (Amazon S3)
Lease: 1 TB/Month 3$ regular; 1.25$ infrequent;
Glacier (archive) 0.70$
Buy: 1 TB 60$
Cost models for Database, management,
storage, network, computation
Data mining/analytics (rapidminer)
Price proportional to number of tuples used
100

25
New Metrics Bid strategies

Cost $$ One time bids

Cost of Computing Present once, exit if it falls below current market price
Trying to figure how much it is going to cost
Bidding for cloud services
Risk: job may not complete,
Computation (AWS, AZURE), B/W, Storage (box, dropbox)
Persistent bids
How to bid the cloud [Liang, Zheng et.al. SIGCOMM 2014]
Need protocols that reduces burden on end-user Resubmit each time period until job finishes
Sliding Scale Hybrid
Important nodes one time bid and slaves persistent
bids (e.g., map-reduce)
Buy Rent
Requires transparent provider pricing
Static, Dynamic Auction Pricing
101 102

Cost continuum tools The Current Network

Routing, management, mobility management,
access control, VPNs,
Can we build tools that can
answer questions like Featur Featu
What is the cost for doing X? e re
What should I own, rent, or bid? Lease Rent Million of 5400 RFCs Barrier to entry
Buy lines
How should I distribute my
Operating
computation for a given cost of source
System
model? code
What is the performance
impact?
Cost vs accuracy, cost vs latency Specialized Packet Billions of Bloated Power Hungry
, cost vs availability Edge Cloud Forwarding Hardware gates
Replace CPU with any other
resource: storage, DB, B/W,
management etc Many complex functions baked into the infrastructure
Edge vs Cloud -- cache to
save dollars OSPF, BGP, multicast, differentiated services,
Traffic Engineering, NAT, firewalls, MPLS, redundant layers,

103
An industry with a mainframe-mentality, reluctant to change
104

Slides from Stanford Site: Nick Mckeown, Martin Casado, Scott Shenker et al,

26
The SDN Approach SDN approach

Separate control from the datapath Conventional SDN

i.e. separate policy from mechanism Application
Network OS Application
ASIC
Datapath: Define minimal network instruction set North bound API

A set of plumbling primitives Application Controller

A vendor-agnostic interface: e.g., OpenFlow Network OS
ASIC South bound API
Control: Define a network-wide OS Openflow, POF, OpenSwitch

An API that others can develop on Switch OS

Application Switch H/W
Network OS
105 ASIC 106

Flow Table Flow Table Entry

Type 0 OpenFlow Switch
Match Priority Counters Instructio Timeouts Cookie
fields ns

Rule Action Stats

Packet + byte counters

1. Forward packet to port(s)

2. Encapsulate and forward to controller
3. Drop packet
4. Send to normal processing pipeline

Switch MAC MAC Eth VLAN IP IP IP TCP TCP

Port src dst type ID Src Dst Prot sport dport
+ mask
At each table , action set is constructed; at the end action set is executed 107
Source: Open flow switch specification-Open Network Foundation

27
Examples Examples

Switching Routing

Switch MAC MAC Eth VLAN IP IP IP TCP TCP Switch MAC MAC Eth VLAN IP IP IP TCP TCP
Action Action
Port src dst type ID Src Dst Prot sport dport Port src dst type ID Src Dst Prot sport dport
* * 00:1f:.. * * * * * * * port6 * * * * * * 5.6.7.8 * * * port6

Flow Switching VLAN Switching

Switch MAC MAC Eth VLAN IP IP IP TCP TCP Switch MAC MAC Eth VLAN IP IP IP TCP TCP
Action Action
Port src dst type ID Src Dst Prot sport dport Port src dst type ID Src Dst Prot sport dport
port6,
port3 00:20.. 00:1f.. 0800 vlan1 1.2.3.4 5.6.7.8 4 17264 80 port6 port7,
* * 00:1f.. * vlan1 * * * * *
port9
Firewall

Switch MAC MAC Eth VLAN IP IP IP TCP TCP

Forward
Port src dst type ID Src Dst Prot sport dport
* * * * * * * * * 22 drop

SDN use cases Boxes, boxes everywhere

Global Traffic Engineering

B4, SWAN
Network function Virtualization Load balancer
Wireless Router
Firewall
Allow multiple functionalities implemented on
Hardware
Made programmable by SDN
Layer 2 Switch
NAT box
Controller

Composed flow tables

111 112

28
SDN issues Research

Correctness, Consistency, Configuration Learning in networks

Network functions
Policy to flow rules
Edge learning
Controller robustness Smart control plane
Security, availability Deep learning
Management Plane Querying vs lookup
Bots for Network management
Network defined Economy
Shared resources
Smart cities
113 114

29