A Case for Rethinking the Internet Architecture: Some Promising Approaches

1
A Case for Rethinking the Internet
ArchitectureSome Promising Approaches
Guru Parulkar http//cleanslate.stanford.edu
2
Executive Summary

• Internet has been a great success
• Has emerged to be a great platform for
  innovations
• It is time to reinvent Internet infrastructure
• Being subject to uses and abuses it is not
  designed for
• Can be a even bigger platform for innovations
• Many promising approaches being put forward
• Infrastructure virtualization and programmability
  key
• CIO organizations have a big role to play
• Active collaboration with infrastructure
  researchers important

3
Internet Has Been A Transformative
Infrastructure Top 20 Engineering Innovations
of 20th Century

• A Century of Innovations(National Academy of
  Engineering)

4
Internet Ossification
Everything on WEB
HTTP
IP
Ethernet
802.11
Satellite
Optical
Power lines
Bluetooth
ATM
5
State of Internet
in the thirty-odd years since its invention,
new uses and abuses, , are pushing the Internet
into realms that its original design neither
anticipated nor easily accommodates. Freezing
forevermore the current architecture would be bad
enough, but in fact the situation is
deteriorating. Overcoming Barriers to
Disruptive Innovation in Networking, NSF Workshp
Report, 05.
6
Internet Architecture Limitations

• Security robustness - to support other critical
  infrastructures
• Control and management
• Addressing, naming (inter-domain) routing
• Mobility of hosts and networks
• Economic viability of different stakeholders
• Scaling of service innovations

7
IP Narrow Waist Not Designed to Absorb
Innovations

• IP narrow waist architecture very successful
• The IP narrow waist decides Internet architecture
  and thus its behavior
• Trust and security
• QoS
• Mobility
• IP narrow waist difficult to change and evolve
• Not designed for evolution
• Does not absorb innovations to the narrow waist
  

8
Internet Architecture Does Not Support Scaling of
Services
Applications
Network Substrate
New service providers have to build and
deploytheir own infrastructure -- tremendous
barrier
Service innovationstarts at the edges
9
Future Internet Research Agenda

• GENI Research Plan
• Compelling comprehensive case for rethinking
  Internet
• 100 page with two years of work
• http//www.geni.net/GDD/GDD-06-28.pdf
• Not universally admired
• One group -- too broad
• Another group -- not broad enough
• Classic science missing -- maybe it should??
• More a reflection of how CS community doesnt
  still know how to champion multiple big agendas

10
Momentum Building Funding

• NSF GENI Initiative
• FIND and other research programs
• GENI as a facility for experimentation
• US multi-agency study on advanced networking
• Europe
• FIRE and other complementary programs
• Japan, Korea,

11
Momentum Building Research

• NSF FIND projects http//www.nets-find.net/
• European FIRE awards
• Many ideas from the past few years
• Stanford Clean Slate Internet Design Program

12
Executive Summary

• Internet has been a great success
• Has emerged to be a great platform for
  innovations
• It is time to reinvent Internet infrastructure
• Being subject to uses and abuses it is not
  designed for
• Can be a even bigger platform for innovations
• Many promising approaches being put forward
• Infrastructure virtualization and programmability
  key
• CIO organizations have a big role to play
• Active collaboration with infrastructure
  researchers important

13
The Stanford Clean Slate Program

• Bring together Stanfords breadth and depth
• Networking, optical communications, wireless,
  access networks, theory, economics, security,
  applications, multimedia, operating systems,
  hardware and VLSI, system architecture,
• Research for long term impact on the practice of
  networkingTwo pronged approach innovations in
  the small and innovations in the large
• Funding
• Industry Collaborators Sponsors Cisco, DT,
  DoCoMo, NEC, Xilinx
• Government Agencies NSF, DARPA,
• Clean slate as a research process, not
  necessarily as a deployment strategy

14
Broad Interdisciplinary Focus
15
Example Projects
Architectural Blueprint of Future Internet?
E2E Secured Network
Open Prog Mobile Internet2020
EASIEnable ScaleInnovations
16
EASI Goals

• Enhancing the Internet to make it an even more
  powerful engine for innovation.
• Two Motivating Cases
• Architectural innovations
• enabling fundamental changes to the Internet
  architecture
• Service-level innovations
• lowering barrier-to-entry for scalable service
  deployment
• Joint project between Stanford and Princeton
  University

17
Future Infrastructure Key Concepts
Slicing, Virtualization, Programmability
18
Future Infrastructure Key Concepts
Slicing, Virtualization, Programmability
19
Future Infrastructure Key Concepts
Slicing, Virtualization, Programmability
20
EASI Infrastructure Observations

• Infrastructure includes
• Computing, storage, packet/circuit switching
  resources
• Slicing, virtualization, programmability
• Across layers layer 1 and up
• Across resource types computing, storage, net
• End-to-end end nodes, enterprise, regional,
  backbone,
• Functionality and performance both important

21
Potential for Innovation

• Allow network providers to support value-added
  services
• Empower end-user innovation
• Allow user-centric innovations to migrate into
  the network
• Allow network architectures to evolve seamlessly
• Provide economic viability to different
  stakeholders
• Democratization of Innovations
• Eric von Hippel

22
EASI in Enterprise
To Internet
23
EASI OpenFlow Network Goals

• A way for researchers to run experiments in the
  networks they use everyday.
• Ease the deployment of GENI in college campuses.
• A pragmatic compromise
• Allow researchers to run experimental
  architectures in their network
• without requiring vendors to expose internal
  workings.

24
No obvious way

• Commercial vendor wont open software and
  hardware development environment
• Complexity of support
• Market protection and barrier to entry
• Hard to build my own
• Prototypes are flakey
• Software only Too slow
• Hardware/software Fanout too small (need gt100
  for wiring closet)

25
Furthermore, we want

• Isolation Regular production traffic untouched
• Virtualized and programmable Different flows
  processed in different ways
• Equipment we can trust in our wiring closet
• Open development environment for all researchers
  (e.g. Linux, Verilog, etc).
• Flexible definitions of a flow
• Individual application traffic
• Aggregated flows
• Alternatives to IP running side-by-side

Flow is a useful granularity for virtualization
26
Network Slicing and Virtualization OpenFlow
Switch
Flow Switch
Open API
Simple Protocol
Secure Channel
SSL
sw
Linux PC

1. Decides which flows to allow
2. Programs flow-table

Flow Table
hw
Packet processing If flow in table, forward
packet Else, send to controller
e.g. 48-ports of 1GE, or wireless access point.
27
Customized Flow Processing
Commercial Switch
Open API
OpenFlow Protocol
SSL
Linux PC

• Line-rate packet processing
• Hardware virtualization
• Congestion Control
• Measurement
• Packet inspection
• New protocols
• ?

Office
Laboratory
Ivo
NetFPGA
28
Server room
controller
PC
29
Deployment at Stanford

• Stanford Computer Science Department
• Gates Building
• 1,000 network users
• Current network
• Catalyst 4k in 22 wiring closets
• 100M Ethernet to desktops
• 2 x 1GE uplinks to basement
• 2 x Catalyst 6500 in basement
• 2 x 10GE uplinks to campus
• Cat 4k coming to end of life
• Expect to announce building-wide OpenFlow shortly

30
The EASI/GENI Big Picture
Enterprise GENIClearinghouse
Experimenter
OpenFlow Network Aggregate
Other Enterprise GENI Components
Controller Aggregate Component Manager (ACM) on
NOX
Desktop, Clusters, Storage, etc.
OpenFlow switches, NetFPGA
31
Nationwide EASI NetworkOver Internet2 Backbone
EASI Enterprise
1GE links
32
Virtual World (VW) Environment

• VW emerging to be another interesting application
  class
• VW is very demanding and will push Internet to
  its limits
• 3D graphics, multimedia streaming and scalability
  critical
• Scalable to 1M concurrent users 50M registered
  users spread over world
• A single VW environment to support many VW
  applications
• Two big components of the environment
• Massive content dissemination and large scale
  simulation
• Neither one supported well on the current Internet

33
VW Environment High Level Structure

• Simulation models the VW for potentially 1M
  users
• Increasingly better fidelity and responsiveness
  are key requirements
• Content provided by providers and 50M
  participants
• Higher quality multimedia and graphics content is
  a must
• Content dissemination provides content to render
  relevant VW for 1M users

34
VW Simulation
A simple model map the world being simulated on
a 2D grid and assign each square of the grid to a
processor
35
VW Simulation on a Slice
Data Center A
Data Center B
Data Center C
Data Center D
VW Simulation on a PVI that is customized for
high end computing with dynamic topologies
36
VW CDN on another Slice
VW CDN on a slice that is customized for caching
at the edges
37
Research Scope
Virtual worlds, gaming, ubiquitous
computing, content dissemination, distributed
storage, sensor integration,
Architecture for a Virtualizable Programmable
Network Substrate
Resource acquisition and configuration, programmin
g models and tools
Security, mobility, robustness, manageability,
scalability
Diverse node configurations data centers,
backbone PoPs, enterprise networks, wireless APs
38
Networking Research and CIO Org

• 1970-80s
• Close collaboration between research groups and
  CIO org
• PhD students worked for CIO and produced some of
  the best research
• 1990s and beyond
• CIO orgs focused on providing service and keeping
  the infrastructure up
• Research becoming too sterile -- not benefiting
  from CIO orgs insights, infrastructure,
  relationships with vendors
• CIO orgs missing the fun and excitement of
  research and innovations
• Everyone loses, rate of innovations slows,
  vendors dominate.
• There are of course exceptions Google and
  Stanford

39
New Opportunities

• Infrastructure virtualization
• Multiple virtual infrastructures on same physical
  infra
• Isolation between production experimental
  traffic/infra
• Production applications can opt-in into
  experimental infrastructure to exploit new
  capabilities
• Stanford and Princeton want to bring this to your
  campus on commercial equipment
• With EASI and OpenFlow
• This is just beginning -- lots of innovations to
  come
• Are you ready? Do you want to enable this?

40
Executive Summary

• Internet has been a great success
• Has emerged to be a critical infrastructure for
  society
• It is time to reinvent Internet infrastructure
• Being subject to uses and abuses it is not
  designed for
• Can be a even bigger platform for innovations
• Many promising approaches being put forward
• E.g., EASI and OpenFlow by Stanford and Princeton
  
• CIO organizations have a big role to play
• Infrastructure virtualization and OpenFlow can be
  key enablers