1

A Scalable Internet Architecture

Nirmala Shenoy, Victor Perotti,Rochester Institute of Technology (RIT)

Technical and socioeconomic studies

Koushik Kar, Aparna GuptaRensselaer Polytechnic Institute (RPI)

Technical and business studies

Murat YukselUniversity of Nevada, Reno (UNR)

Technical studies

2

Introduction

• Small Grants Exploratory Research (SGER)– High risk research idea

• Revolutionary design to a clean slate future Internet • But – based on well proven concepts

OUTCOMES

3

A Scalable Internet Architecture• Technical

• Architectural studies – includes addressing, protocols• Support for wireless networks and roaming users

– Transition • Socioeconomic & Business impacts

• Stakeholder incentives– Transition

• 6 months into the new grant

4

Technical Questions

– Do we need 2 addresses? Logical and PhysicalObservation – global Logical addresses - look up and

extensive routing– 32 not enough, 128 not wireless friendly

Flexible - tailored to requirement, easy transition

– Current Internet – optimize on meshing When to mesh and how much?Combination of hierarchy and mesh structures

– Given above- relook at the protocol stack?

Addr

essi

ng

Topological consideratio

n

Prot

ocol

s

The above should be ISP friendly. Easy transition

5

Architectural studies

AddressingStructural topology

Wireless networks support

6

Addressing in the ISP topologyHigh levelTypical -

RIT RPI

BB2BB1

BB3

DR1

DR2 DR3

DR4

AR1 AR2 AR3AR4

One POP

CLOUDCLOUDCLOUD

Tier 1

Tier 2

Tier 3

UofN

Tier based addressing - inter and intra cloud

Tier 1prefix

Tier 2 prefix

1. 1

2. 1: 1

BB cloud identifier

DR cloud identifier

BB Cloud identifier

1.1:1 BB1 BB3

BB2

DR4

DR3DR2

DR1

1.1:2

1.1:3

2.1:1:1

2.1:1:2 2.1:1:3

2.1:1:4

Tier 3 prefix 3. 1: 1: 1AR4AR3AR3AR1

3.1:1:1:1 3.1:1:1:2 3.1:1:1:3 3.1:1:1:4

Tier 4 prefix 4. 1:1:1:1??

Flexible Addressing - Wireless friendly

Rensselaer Polytechnic Institute

Rochester Institute of Technology

2.1:2:x

8

Flexibile Addressing Scheme

Tier field Length field Address field

6 bits 2 bits 4- 12 bits

6 bits – 64 tiers“01”“10”“11”“00” Special handling

4 bits 16 systems8 bits 256 systems12 bits 4096 systems

Length field Address field

2.1:1:4

tier 1st address 2nd address 3rd address

000010 01 0001 01 0001 01 0100

9

Flexibility in Addressing Scheme

• Faster forwarding between tiers – • Switch on tier field – UP, DOWN, • Same level –Mesh as required and route - maximum 4096• Distribute routing load to within tiers

• Address length depends on tier level – (no fixed size)• Addresses will never run out• 00 - Special addressing – wireless networks, roaming user• Flexibility – nested addressing

10

Nested AddressingCloud ID 1.1

Cloud ID– 2.1:1

Cloud ID– 3.1:1:1

Tier 1 – 1.1

Tier 2 – 2.1:1

Flexibility at tier levelInternal addressing structure

Private, partly private/publicDNS to locate DNS hierarchy

domain name service /tier

BB1 BB2

BB3

DR1

DR2 DR3

DR4

AR3AR4

Client networkSmall ISP

BB1BB2

OR1 OR2

Global ID– 3.1:1:1:x

DNS

DNS

DNS

11

Special Handling Length field = 00 – No forwardingAddress field

– 00 – Last address field– 01 - Roaming user– 02 - Proprietary addressing

• IPv4• IPv6?

– 03 – sensor networks– 04 – user requires an address– 05 – user requires some service

Length field

2 bits

01 – 4 bits10 – 8 bits11- 12 bits

00- special handling

12

Roaming User

MN

Request – length=00, Afield=01

AAA server

Address assigned – 3.1:1:1:1:x

MNRequest – length=00, Afield=01

Address assigned – 3.1:1:1:2:x

NO HOME ADDRESS CONCEPT – LOCATE BY NAME SERVICE ONLY

BB1 BB2

BB3

DR1

DR2 DR3

DR4

AR1 AR2 AR4

Cloud ID 1.1

Cloud ID– 2.1:2 Cloud ID– 2.1:1

Cloud ID– 3.1:1:1 Cloud ID– 3.1:2:1

3.1:2:1:4

3.1:1:1:2 3.1:1:1:1

AAA server

AAA server

TIER LEVEL DISTRIBUTED AAA SERVERs

13

The Protocol StackApplication with reliable

transmission if requiredPort Services

Switch up/down, Intra-tier routing

Medium Access control

Physical layer

Max routing table size -4096Tier 1 ISPs - around 10 entriesImpact on all optical networks

Inter ISP agreements (RPI &UNR)

ISP - transit services (RPI & UNR)

14

Architectural studies - Progress • Proposed architecture, addressing developed

in Opnet• Imported AT & T topology from Rocketfuel• Imported to Opnet for simulation studies• Imposing hierarchy in the ISP topology

– Cytoscape based optimization study

15

AT & T Seattle POP

Rocket fuel – Cytoscape (architectural studies)

16

Address Length DistributionAT&T (USA)

545, 5%

3051, 26%

187, 2%213, 2%

501, 4%124, 1%

1709, 15%

2230, 20%

2843, 25%

12bit16bit18bit22bit24bit26bit28bit32bit36bit

0

4

8

12

16

20

24

28

32

36

40

1 792 1583 2374 3165 3956 4747 5538 6329 7120 7911 8702 9493 10284 11075

17

Wireless Network-Roaming studies

• Studying seamless roaming scenarios• Addressing during roaming• Mobility Anchor points concept

– Common prefix– Optimization

• AAA service• DNS service

18

Transition studies (Technical)• Inherent in the solutions

– Structural topology– ISP topologies start point– Addressing – ISP and wireless friendly

• backward compatible – What is different?

• The structure is not an overlay– Effect of underlying structure is still there– Virtual effect, complex

• Solution can be implemented at layer 2– Impact the underlying structure

• Modification to MPLS – HLSP (hierarchical Label switching) • Label stacking – hierarchy can be introduced

19

Test Beds

• Test bed of 12 Linux systems– IP applications

• MPLS test bed - transition• MPLS based testing at RIT network

20

11 12

23

24

31

21

22

35

34

36

32

3337

21-32 / 32

21-22-33 / 22 – 33 / 33

21-22-34 / 22-34 /34

21- 12- 24- 37 / 12 -24-37/ ….

21-11-23-35 / 11-23-35/ ……

21 – 12- 24- 36 / 12-24-36/ …

Tier 1

Tier 2

Tier 3

1

12

123

124

1231

12311124 1

122

122112211

122 1

11111

1121121

1111

1112

11211

11121

11121211112

111121

1111212

RIP ?OSPF ?DIJSKTRA’S

MULTI MESHED TREEALGORITHM

22

Socioeconomic and Business Impact studies

The Socioeconomic Piece: Towards Understanding the Market Context for Network Innovation

• Initial Business Industry Analysis• Buyer Power• Language check: Deployment ≠ Adoption

Which forces are prevalent in the Internet Industry?

Traditional Business Analysis (ala Porter) looks at 5 Forces in an Industry.

Collectively, these forces inform the business opportunities available in this Industry

Note: The threat of New Entrants and Supplier Power are the key Drivers according to Datamonitor 2007

Which entities are capturing value?

Shawn O'Donnell’s 2002 Economic Map of the Internet is being enhanced to reflect the state of the 2009 Internet Industry.

An Economic Value Map can help to identify key influencers in the Industry.

But what else is happening?

“Web could collapse as video demand soars” –Daily Telegraph 4/28/2008

“Study: 44% Of Internet Traffic Is Peer-to-Peer” Multichannel News 6/23/2008

“NebuAd loses CEO, business model in wake of tracking furor” Ars Technica 12/5/2008

susanbeebe RT @Dan_Agnew: RT @LeeDrake: I am putting together a website to try to attract VerizonFIOS to Rochester. #TWC needs some real competition about 8 hours ago from TweetDeck

Traditional analysis does not capture the new importance of users

• Scott Jordan’s 4 Tenets – ISP adoption driven by consumer “product”

(example VOIP or Video)• Fixing problems versus enabling features• “Glacial” adoption of IPv6

Science Engineering Economics Business Here Be Dragons

Know the ConsumersThe Groups that will influence the adoption of a new Internet

technology

0.00%

10.00%

20.00%

30.00%

40.00%

50.00%

60.00%

70.00%

80.00%

90.00%

ISPs InternetCompanies

Netw orkHardw are

Vendors EndUser Govt Other Group

Groups

Influ

ence

We are conducting a study of Internet stakeholders to evaluate the adoption of a new Internet infrastructure. Initial 90 respondents from NANOG 2009

And further

• Clemon’s concept of Informedness alters traditional thinking on Product, Promotion, Pricing(2008).

• How might we designing “Products” and/or Incentives for various target Customers (ISP, CIO, End User)

• Contract Switching…coming up next

30

QUESTIONS ?