CogNet - Cognitive NetworkingCogNet - Cognitive Networking

NSF NeTS/FIND (Future Internet Network Design)Collaborative ProjectRutgers University

University of KansasCarnegie Mellon University

CogNet in PerspectiveCogNet in Perspective• GENI (Global Environment for Network

Innovations)• Global experimental facility that will foster

exploration and evaluation of new networking architectures (at scale) under realistic conditions

• Major infrastructure, expected to be $367 million

• FIND (Future Internet Network Design)• Requirements for global network of fifteen

years from now - what should that network look like and do?

• How would we re-conceive tomorrow's global network today, if we could design it from scratch?

• Innovative ideas in broad area of network architecture, principles, and design

• Research projects expected to be funded at $20 million per year in progressive phases

• Provides experiments and architectures that will be pursued on the GENI infrastructure

Wireless Networking ChallengesWireless Networking Challenges

Why is wireless networking hard?

• Mobility is inherent with untethered

• Resources are constrained• Spectrum “scarcity” → bandwidth & delay issues

• Environment changes• Mobility → different surroundings (indoor, urban, rural)

• Varying physical properties• Wireless communication path changes over time

Cognitive / Agile Radio PlatformsCognitive / Agile Radio Platforms• Flexible in RF carrier frequency (~0 - 6 GHz)• Flexible in bandwidth (several 10’s MHz)• Flexible in waveform

• A/D and D/A driven• Generated/processed by programmable DSP and/or FPGAs

• Dials to observe• Traffic characteristics measured at network

layer• Error rate & characteristics (BER and

distribution)– MAC layer per packet error information– Network and transport layer per flow

correlations• Receive characteristics

– Physical layer – signal strength, interfering signals, background noise

– MAC layer – transmit power, antenna in use

• Knobs to influence• Physical layer

– Frequency & bandwidth– Transmit power– Beam width & direction– Data rate, code, & chipping rate

• MAC protocol– FEC strength– Retransmit scheme– MTU size– Encryption & parameters

• Network layer– Routing protocol– Addressing plan(s)– ACLs

• Interface framework with a flexible, usable set of scalable parameters• Adapt to resource constraints, environment, varying physical conditions, application

AD8347 80AD8349 150ADF4360 4 @60mA 240ADF4113 15SMV3300A 20FOX801BE-160 5BGA2031 70MGA-83563 200MGA-82563 100MGA-545P8 140MGA-86576 2 @25mA 50HMC488MSG8 2 @50mA 100ERA-1SM 40

+3.3 +5

770mA 440mA

3 dBPOWERDIVDER

SMV3300A

LO 3

LNA

BASEBAND I

BASEBAND Q

SPI Bus

5 GHz RF PCB Block Diagram

5.250-5.850 GHz

AD8347

I-Q DEMODULATOR800 MHz - 2.7 GHz

GAINCONTROL

I DET Q DET

LNA

1.850-2.450 GHz

AD8349I-Q MODULATOR700 MHz - 2.7 GHz

090

BASEBAND Q

BASEBAND I

LR I

PA

ADF4360-1

ADF4360-2

TX LO 2

3.4 GHz

16.0 MHz

TCXOFOX801BE-160

0

90

D. DePardo

1 1

27 JULY 05

BGA2031Gain = +23dB@1.9 GHz (2.7

G = 56dB@1.9 GHzP1dB = +13dBm@1.9 GHz

VCT RL

)

MGA-86576Gain = +19dB@6.0 GHzP1dB = +4.3dBm@6.0 GHzNF =1.8dB@6.0 GHz

HMC488MS8GCLoss = 10 dBP1dB = +8 dBmLO Drive 0dBm

MGA-83563Gain = +17dB@6.0 GHzP1dB = +15dBm@6.0 GHzP = +18dBm@6.0 GHzSAT

Lumped Microstrip

REF CLK OUT

LR I

MGA-82563Gain = +10dB@4.0 GHzP1dB =+17dBm@4.0 GHz

(-100dBm)

+8dBi

-4dB +19dB

(-77dBm)

-5dB

-3dB

+19dB

(-66dBm)

(+5dBm)

-5dB+9dB

(+3.5dBm)

-2dB

(+3.5dBm)

-10dB -5dB

(-81dBm)

(+3dBm)

-5dB+9dB

(+7dBm)

-10dB-5dB+17dB

(+9dBm)(+21dBm)

(+25dBm)

+8dBi

JumperSelect

BW=30 MHz

BW=30 MHz

-4dB

(+17dBm)

ADF4360-1

ADF4360-21.850-2.150 GHz

RX LO 1

AD56016 BIT DAC

RX IF GAINCONTROL

AD56016 BIT DAC

TX IF GAINCONTROL

1

0-30dB RXAttenuation

Control

Act ive TX Antenna Module

Act ive RX Antenna Module

2.150-2.450 GHz

1.850-2.150 GHz

2.150-2.450 GHz

3.4 GHz

1.850-2.450 GHz

5.250-5.850 GHz

(-3dBm)

OPTIONALFor use with passive

antennaMGA-545P8Gain = +11.5dB@5.8 GHzP1dB = +21dBm@5.8 GHzP = +22dBm@5.8 GHzSAT

ERA-1SMGain = +6dB@6.0 GHzP1dB = +12dBm

(-8dBm)(-2dBm)(+15dBm)

ADF4113

MC68HC08Microcontroller

I²C

DPB Input

Cognitive NetworkingCognitive Networking

Cognitive (from Wikipedia) – applying the experience gathered in one place by one being

to actions by another being elsewhere

• Developing experimental protocol stack for cognitive networks, not just cognitive radios

• Scalable autoconfiguration & network management• Dynamic network layer supporting tailored functionality

(IP, group messaging, rich queries, etc.)• Builds on the foundation of cognitive radios (e.g., Rutgers

/ GNU Radio, KU Agile Radio), but extends it further up the protocol stack, and explores across stack

CogNet VisionCogNet Vision

Network Layer OverlaysNetwork Layer Overlays

Core Network

Supernode(mobile or fixed)

Mobile Nodes

Overlay 1

Overlay 2

Network Layer Innovations ExampleNetwork Layer Innovations Example

• Sensor networks with resource constraints• Size, weight, power limits → bandwidth, processing power limits

• Traditional IP networking not the best answer• Substantial overhead• Unnecessary communications

• Innovations• Lightweight network-layer protocols• Operations on data flow – e.g., nodes

do not forward sensor data valuesthat are unchanged

• Gateways

Wireless Sensor Network

Network Layer OverlaysNetwork Layer Overlays• Structured and unstructured P2P, DHT• Services may map better to particular overlays

• Search, distributed file storage, load balancing, multicast messaging

• Overlay typically denotes an application layer network of semi-persistent links between participating nodes, that is used to forward messages between the distributed application elements

• Can also use them for Layer 3 forwarding• Inspired by M. Caesar, M. Castro, E. Nightingale, G. O'Shea and A.

Rowstron, "Virtual Ring Routing: Network routing inspired by DHTs", Sigcomm 2006, Pisa, Italy, September 2006, http://research.microsoft.com/~antr/VRR.htm

• Explore how having tailored layer 3s, (IP, range-based overlays, multicast optimized overlays, etc.) may impact end-to-end network architecture for interoperating cognitive wireless subnets and the future Internet

Network Layer OverlaysNetwork Layer Overlays

• Research topics• How to use, position, and discover routers between the overlays

themselves, and the Internet• How applications can decide which network layer to use

– Legacy approach manipulating resolver libraries– New approach by applications aware of the Global Control

Plane (GCP)• Implement multiple new network layers (linux kernel changes for

experiments, implement within ns2 for simulations, and explore if common code can be leveraged)

• Ideally explore performance tradeoffs (more overhead, etc. versus better utilization, etc.) in simulation and real cognitive radio network (KUAR or Rutgers/GNU Radio)

Routing and PlatformsRouting and Platforms• XORP

• Routing protocol implementations, extensible API, and configuration tools• IPv4 and IPv6 – BGP, RIP, PIM-SM, and IGMP/MLD

• Routing Control Platform (RCP)• Computes BGP routes for the routers in an AS• Incorporates complete routing information & network engineering

• Provides basis for experimentation with interdomain routing• Inter-overlay – between overlays on same network• Gateways (supernodes) between networks

Source: XORP Design Overview

Source: M. Caesar, et al, "Design and implementation of a Routing Control Platform“, NSDI 2005

XORP

Network Management ArchitectureNetwork Management Architecture

• Global Control Plane

Network Management Information

Exposed ViaManagement AndService Discovery

Overlays

Network Management Information

Exposed ViaManagement AndService Discovery

Overlays

Network Wide Cross Layer Interaction

Cognitive Radio & MAC Layers

Local Instance

of theGlobal ControlPlane

Network Management Information

Exposed ViaManagement AndService Discovery

Overlays

Applications

Cognitive Networking Layer

IPRich

QueryGroup

Messaging

Node 1 Node 2 Node 3

Global Control PlaneGlobal Control Plane• Implement overlay for inter-node transmission of control

plane data (position, capabilities, errors, signal strength, etc.)

• Implement parts of local node control plane necessary to perform network layer research• Primarily application and layer 3• CMU/Rutgers will implement layer 1 & 2 for their radios • KU may implement local node control plane layer 1 & 2 for KUAR

• Explore performance – analytically & experiment• Overlay overhead versus better data for cognitive decisions using local

cross-layer and global cross-network information• Assertion - make better cognitive decision knowing local node

information and receiving node environments as well as details above any intermediate hops

CogNet SummaryCogNet Summary• Developing experimental protocol stack for cognitive networks, not just cognitive radios

• Scalable autoconfiguration & network management

• Dynamic network layer supporting tailored functionality (IP, group messaging, rich queries, etc.)

• Building on foundation of cognitive radios (e.g., Rutgers & GNU Radio, KU Agile Radio), but extends it further up the protocol stack, and explores across stack

CogNet - Cognitive NetworkingCogNet - Cognitive Networking