The Internet, or Something BetterAn Update on the NetSE Council

Ellen W. ZeguraChair, NetSE Council

School of Computer Science, Georgia Tech28 October 2008

Pictures with captions

Acholi children in an InternationalDisplaced Persons camp in Kitgum

(wikipedia)

The Internet topology(wikipedia)

Current Internet

0102030405060708090

Filtering RouteLeaks

RouteHijacks

RouteInstability

RoutingLoops

Blackholes

Occ

urre

nces

ove

r 10

Yea

rs

1995-1997 1998-2001 2001-2004

Intellectual Space(not to scale)

Network Design and Engineering

Economics (Behavior)and Networks

Societal Values[Are we missinganything?]

Goal: Understand how to design, engineer and operate “better” networks

Network Science

Theoretical CS

NetSE: Community effort

• Ellen Zegura, GT, chair• Joe Berthold, Ciena• Charlie Catlett, Argonne• Mike Dahlin, UT Austin• Chip Elliot – GPO (ex-officio)• Joan Feigenbaum, Yale• Stephanie Forrest, UNM• Roscoe Giles, Boston Univ

• Michael Kearns, UPenn• Ed Lazowska, Washington• Peter Lee, CMU• Helen Nissenbaum, NYU• Larry Peterson, Princeton• Jennifer Rexford, Princeton

Mission: The primary mission of the Network Science and Engineering (NetSE) Council is to articulate a compelling research agenda for Network Science and Engineering, including inter-related theoretical, experimental and societal aspects.

Additions pending – workshop leaders, other enthusiasts

Workshops• Science of Network Design

– John Doyle, CalTech/NSF– John Wroclawski, ISI– July 29 and 30, southern CA

• Behavior, Computation and Networks– Mike Kearns, U Penn– Colin Camerer, CalTech– July 31 and August 1, La Jolla

• Network Design and Societal Values– David Clark, MIT– Helen Nissenbaum, NYU– September 24-26, Washington DC

Meetings

• Smaller than workshops• Extract/expand on more well trod areas

• Theory and Network Design– John Byers (BU), Joan Feigenbaum (Yale), Ellen Zegura (GT)– June 11, Boston

• Network Design and Engineering– Nick Feamster (GT), Amin Vahdat (UCSD), David Andersen

(CMU), Mike Dahlin (UT Austin), Jen Rexford (Princeton), Craig Partridge (BBN), David Clark (MIT), Dmitri Krioukov (CAIDA), NSF folks, GPO folks

– August 17, 18, Seattle at SIGCOMM

Timeline

• June-Sept 2008 – elaborate the space– workshops (3)– meetings (2)

• Oct 2008– draft research agenda completed– incl. recommendations on how to advance agenda

• Nov 2008– collect feedback (from few then many)

• December 2008– finalize research agenda

Brief overview of workshops

• Science of network design

• Behavior, computation and networks

• Network design and engineering

• Network design and societal values

Science of Design Workshop

SoD Themes

• Models and model validation – from self-referential to observational to

generative (predictive)– semantics

• Architecture derivation– from optimization frameworks– combined with past examples

• Local-global interactions and design

Example: Global Results from Local Actions“Design by Constraints” (JTW)

Low degree mesh-like core• High performance and

robustness• Efficient, economic• From “random” generator,

low probability, but• Like real Internet

High degree hub-like core• Poor performance and

robustness• Wasteful, expensive• From “random” generator,

high probability, but• Unlike real Internet

100 101 102 103100

101

102

power-law degrees

Two “Internet Topologies”; same power law parameters..

Why is this? (JTW)

• This is surprising to many in network science• This is not surprising to most Internet engineers• What’s going on?

SOX

SFGP/AMPATH

U. Florida

U. So. Florida

Miss StateGigaPoP

WiscREN

SURFNetRutgers

MANLAN

NorthernCrossroads

Mid-AtlanticCrossroads

Drexel U.

U. Delaware

PSC

NCNI/MCNC

MAGPI

UMD NGIX

DARPABossNet

GEANT

Seattle

Sunnyvale

Los Angeles

Houston

DenverKansas

CityIndian-apolis

Atlanta

Wash D.C.

Chicago

New York

OARNET

Northern LightsIndiana GigaPoP

MeritU. Louisville

NYSERNet

U. Memphis

Great Plains

OneNetArizona St.

U. Arizona

Qwest Labs

UNM

OregonGigaPoP

Front RangeGigaPoP

Texas Tech

Tulane U.

North TexasGigaPoP

TexasGigaPoP

LaNet

UT Austin

CENIC

UniNet

WIDE

AMES NGIX

PacificNorthwestGigaPoP U. Hawaii

PacificWave

ESnet

TransPAC/APAN

Iowa St.

Florida A&MUT-SWMed Ctr.

NCSA

MREN

SINet

WPI

StarLight

IntermountainGigaPoP

Abilene BackbonePhysical Connectivity

Start with an engineeredbackbone…

Add gateway routers and end systems consistently with technological constraints on these routers and systems…

Get topology [synthesized or real] with high throughput, efficiency, economy

Design by constraint (JTW)

• The desirable topology is due to both– Classical engineering– Local constraints shaping global results

• To be fair, perhaps somewhat by accident..

• The key question: can we do it on purpose?– Design, not of the complete system, but of

components from which systems with desired properties will come forth?

– Formalization of methods for this class of design

Behavior, Computation and Networks Workshop (MK)

• Bring together researchers with behavioral and computational interests

• Introduce a behavioral component to Algorithmic Game Theory (AGT)

• Examine how AGT can inform/shape behavioral experiments

• Discuss opportunities and challenges for larger-scale human subject experiments– CS/AGT strength: system design, theories of scaling,

network models– Behavioral Game Theory strength: experimental

design and methodology, network models

Workshop Topics (MK)

• How do computational considerations modify BGT?• How do behavioral considerations modify AGT?• How can we “scale up” behavioral experiments?

– examination of (social) network effects– use of web/Internet technology– development of shared experimental infrastructure/platforms

• If there were no technology/methodology limits, what experiments would we do?

• How can we mix human subjects and artificial agents in interesting ways?

• What are the appropriate (statistical) models for collective behavioral data?

Better: Human-network understanding (MK)

Network Design and Engineering Meeting Themes

• Complexity of Requirements– Exemplar question: Should a future Internet achieve

five 9’s reliability?

• Abundance of Technology– Exemplar question: How do we enable on-the-fly

composition of network protocols across heterogeneous devices?

• Need for Experimentation– Exemplar question: What experimentation platforms

are most useful?

Example: Heterogeneous Mobility

Node Density

“Mob

ilit

y”

High

Low

High

Space Paths

Low

No (Space/Time)Paths

Space/Time Paths

Hybrid Environments

Implications for Routing

DTN ROUTING

MANET ROUTING

SPARSE NET ROUTING

Network Design and Societal Values Workshop

• Personal – a fascinating experience of following what someone is saying right up until they say a word that cannot possibly fit into the meaning you had previously been assuming

Network Design and Societal Values Issues and Themes

• Example issues:– Security, privacy, identity, trust– Openness– Conditions and motivations for participation

• Example, cross-cutting themes:– Visibility and transparency – Incentives– Networks as experimental environments

Better: Internet penetration

Better: Human-network interface

• Ask people to sketch their networks, these are two renditions of the same network. One is technically right, but is the other “wrong”. The other shows the network embedded in local familial practices... [credit Beki Grinter, Keith Edwards]

Some research challenges

• Human-network interface• Materially more secure network w/out lockdown• Networking/computing at the margins• Mobility• Circuit->Packet->Network switching• Science of network design (not your physicist

mother’s network science)• New technologies – quantum, mcore,

programmable down to the waveform

Thank you!

Questions and feedback welcome