Luxbg fringe

The wireless fringe

Uni.lu, December 2012

Pascal Thubert (Cisco Systems)

© 2012 Cisco and/or its affiliates. All rights reserved. UnclassifiedUni.lu 2

Wireless: the evolution trait

Cheap Install

Deploying wire is slow and costly

Global Coverage

From Near Field to Satellite via 3/4G

Everywhere copper/fiber cannot reach

Cheap multipoint access

New types of devices (Internet Of Things)

New usages (X-automation, Mobile Internet)


Agenda

The Fringe of the Internet

The Route-Over Fringe

The Mesh-Under Fringe

The Overlay Fringe

The RPL Fringe protocol

© 2010 Cisco and/or its affiliates. All rights reserved. 4UnclassifiedBRKEWN-3012

The Fringe of the Internet


The routing Infrastructure today

The Internet

Fully engineeredHierarchical, Aggregations, ASs, Wire links

Fully distributed StatesShows limits (BGP tables, addr. depletion)

Reached adult size,

mature to aging

IntranetsSame structure as the Internet

Yet decoupled from the InternetNAT, Socks, Proxies

First model for Internet extension


L2 mesh UnderMulti-hop Public Access Points, Proprietary mission specific products

L3 Route OverMigration to IETF Protocols (RPL)Internet of Things (IOT) Machine to Machine (M2M)

Mobile OverlaysGlobal reachability

Route Projection

Fixed wired

Infrastructure

56

78

CB

1

32

A4

A’s

Home

B’s

Home

MANET

MeshThe Fringe DOES NOT LEAK

into the Routing Infrastructure

NEMO

The emerging Fringe of the Internet

Edge


The Route-Over Fringe


Emergency

HotSpot

(roadside)

SOS

!

Mobile Router

Mobile Router

IPv6

IPv6

IPv6

Swarming


Sensor Dust

“Sensor dust” spread over a territory

Sensors assume a fixed arbitrary geographical distribution

Numerous sensors with limited capabilities (battery …)

A limited number of relays (MR)

MRs run an SGP (RPL)

2 to 3 uplinks (MR with backhaul capability)


Fleet

Global motion plus relative mobility

Managed hierarchy over dynamic topology

Secured uplink to base

Dark Zone coverage and range extension (nesting)

TLMR


Internet

MR1

CN

HA1

HA1: HA of MR1

HA2: HA of MR2

HA-VMN: HA of VMN

CR: Correspondent Router

HA2

VMN

MR2

HA

CN2

CN1

Nested NEMO Route optimization


Forming the nested NEMO

Attachment selection fixed vs Mobile Router

Preventing loops in nested NEMOtopology

Optimize Default Route selection, shallow trees/DAGs

Fast reconfiguration upon movements

Potential attachment

Based on RA reception

Internet


MANEMO

Couple IPv6 global mobility with RPL

Mobility from NEMO, LISP, other…

Minimum set of rules for all MRs

Attach whenever possible

Generic RPL loop avoidance

Delay Attachment by target depth

Individual attachment

May use different OF

Common metrics

Ordered Default Router List

for fast switching / recovery

MONAMI -> InstancesDefault route

Kept in DRL

Dropped

Internet


The Mesh-Under Fringe


Monitoring and Automation

Research & Discovery

Healthcare

Energy

Efficiency

Predictive maintenance

Industrial Automation

Smart

Home

Defense

Asset

tracking

Intelligent Building

Smart CitiesSmart Grid

Car 2 Car

Agriculture


ISA100: Wireless Systems for Industrial Automation

ISA100.11a industrial WSN

Wireless systems for industrial automation

Process control and related applications

Leverages 802.15.4 + IPv6

Link Local Join process

Global Address runtime

6LoWPAN Header Compression

Yet specific routing and ND

Next: Backbone Router

ISA100.15 backhaul


Internet

ISA100.11 / ISA100.15 reference model

ISA100.11a

Backbone

Router

System

Manager

Gateway

(ALG)

Plant

network

Security

Manager

ISA100.15

Backhaul

Router


What’s a Backbone Router?

Common ND based abstraction over a backbone

Scales DAD operations (distributes6LoWPAN ND LBR)

Scales the subnetwork (high speed backbone)

Allows interaction with nodes on the backbone or in other subnets running different operations

http://tools.ietf.org/html/draft-thubert-6lowpan-backbone-router












The RPL (pronounced ripple) Fringe Protocol


Routing With RPL

Dynamic Topologies

Peer selection

Constrained Objects

Fuzzy Links

Routing, local Mobility

Global Mobility

New Radios issues: Addressed in RPL ?


RPL key concepts

Minimum topological awareness

Data Path validation

Non-Equal Cost Multipath Fwd

Instantiation per constraints/metrics

Autonomic Subnet G/W Protocol

Optimized Diffusion over NBMA


Controlling the control … by design

Distance Vector vs. Link State

Knowledge of SubDAG addresses and children links

Lesser topology awareness => lesser sensitivity to change

No database Synchronization => Adapted to movement

Optimized for Edge operation

Optimized for P2MP / MP2P, stretch for arbitrary P2P

Least Overhead Routing Approach via common ancestor

Proactive vs. Reactive

Actually both with so-called P2P draft

Datapath validation


Datapath Validation

Control Information in Data Packets:

Instance ID

Hop-By-Hop Header Sender Rank

Direction (UP/Down)

Errors detected if:

- No route further down for packet going down

- No route for packet going down

- Rank and direction do not match


In the context of routing, a DAG is formed by a collection of vertices (nodes) and edges (links), each edge connecting one node to another (directed) in such a way that it is not possible to start at Node X and follow a directed path that cycles back to Node X (acyclic).

Directed Acyclic Graph for NECM


Generic Rank-based Loop Avoidance

1) A root has a Rank of 1. A router has a Rank that is higher than that of its DAG parents.

2) A Router that is no moreattached to a DAG MUST poisonits routes, either by advertising an INFINITE_RANK or byforming a floating DAG.

3) A Router that is already part of a DAG MAY move atany time in order to get closerto the root of its current DAGin order to reduce its own Rank

4) But the Router MUST NOT move down its DAG – but under controlled limits

whereby the router is allowed alimited excursion down

5) A Router MAY jump from itscurrent DAG into any different DAG at any time and whateverthe Rank it reaches there,unless it has been a member of the new DAG in which case rule

4) applies


Global versus Local Repair

: : A new DODAG iteration

Rebuild the DAG … Then repaint the prefixes upon changes

A new Sequence number generated by the root

A router forwards to a parent or as a host over next iteration

: find a “quick” local repair path

Only requiring local changes !

May not be optimal according to the OF

Moving UP and Jumping are cool.

Moving Down is risky: Count to Infinity Control


Objective Function

Extend the generic behaviorFor a specific need / use case

Used in parent selectionContraints

Policies Position in the DAG

Metrics

Computes the Rank incrementBased on hop metrics

Do NOT use OF0 for adhoc radios!

(OF 0 uses traditional weighted hop count)


Routing Metrics in LLNs

Node Metrics Link Metrics

Node State and Attributes Object

Purpose is to reflects node workload (CPU,

Memory…)

“O” flag signals overload of resource

“A” flag signal node can act as traffic

aggregator

Throughput Object

Currently available throughput (Bytes per

second)

Throughput range supported

Node Energy Object

“T” flag: Node type: 0 = Mains, 1 = Battery, 2 =

Scavenger

“I” bit: Use node type as a constraint

(include/exclude)

“E” flag: Estimated energy remaining

Latency

Can be used as a metric or constraint

Constraint - max latency allowable on path

Metric - additive metric updated along path

Hop Count Object

Can be used as a metric or constraint

Constraint - max number of hops that can be

traversed

Metric - total number of hops traversed

Link Reliability

Link Quality Level Reliability (LQL)

0=Unknown, 1=High, 2=Medium, 3=Low

Expected Transmission Count (ETX)

(Average number of TX to deliver a

packet)

Link Colour

Metric or constraint, arbitrary admin value

For YourReference


Simulation Results

Traffic Control

Traffic Holes – Global Repair only

Routing Table Sizes

For YourReference


Example radio connecticity

Reachability imposed by L2 radio

Variable, almost per packet links


Example radio connecticity

At a given point of time connectivity is

(fuzzy)

Radio link


Clusterhead

1st pass (DIO)

Establishes a logical DAG topology

Trickle Subnet/config Info

Sets default route

Self forming / self healing

2nd pass (DAO)

paints with addresses and prefixes

Any to any reachability

But forwarding over DAG only

saturates upper links of the DAG

And does not use the full mesh properly

Applying RPL

Link selected as parent link

Potential link

Clusterhead0

1

11

4

4

4

46

3

3

3

3

3

2

2

2

22

2

5

55


Clusterhead

5

4

4

A second root is available

(within the same instance)

The DAG is partitioned

1 root = 1 DODAG

1 Node belongs to 1 DODAG

(at most, per instance)

Nodes may JUMP

from one DODAG to the next

Nodes may MOVE

up the DODAG

Going Down MAY cause loops

May be done under CTI control

Multiple DODAGs within Instance

Link selected and oriented by DIO

Potential link

0

1

3

1 1

2

2

2

22

3

3

3

3

3

3

2

4

4

5

0

65

4


Clusterhead0

1

11

2

2

2

22

3

3

3

3

3

3

2

3

5

4

4

4

4

Clusterhead

Multiple Instances

Running as Ships-in-the-night

1 instance = 1 DAG

A DAG implementsconstraints

Serving differentObjective Functions

Using different metrics

Forwarding along a DODAG (like a vlan) Constrained instance

Default instance

Potential link

A


Clusterhead

5

Clusterhead0

1

11

2

2

2

22

3

3

3

3

3

3

2

3

5

4

4

4

4

Applying ARCs

ARC scoped Advertisements

SubDAG via its root

Adv Scope == ARC

Normal DIO up.

Now forwarding over DAGAND ARCs

Reduces congestions of upper links of the DAG

Still LORA for P2P

IGP subarea (bidirectional)

Link selected and oriented by TD

Potential link


Summary

DV, ORA P2MP/MP2P, LORA P2P

Objective Functions, Metrics

Controlling the control

NECM Directed Acyclic Graphs

Trickle and Datapath validation

Local and Global Recovery

N/A

Dynamic Topologies

Peer selection

Constrained Objects

Fuzzy Links

Routing, local Mobility

Global Mobility

New Radios issues: Addressed in RPL by:


Next steps…

Reactive model (already started, aka P2P)

PCE (ala TSMP/ISA100.11a/WiHART)

DAG limitations

Sibling routing

Other resilient schemes (ARCs)

Stimulated updates (lookup)

Asymmetrical links

Multi-Topology routing and cascading

© 2012 Cisco and/or its affiliates. All rights reserved.IoT6 38Unclassified

“We might be at the eve of pervasive networking, a vision for the Internet where every person and every device is connected to the network in the ultimate realization of Metcalf's Law.”


BACKUP Material


The Radio Enabler


Wireless: the evolution trait

Cheap Install

Deploying wire is slow and costly

Global Coverage

From Near Field to Satellite via 3/4G

Everywhere copper/fiber cannot reach

Cheap multipoint access

New types of devices (Internet Of Things)

New usages (X-automation, Mobile Internet)


Dynamic topologies

No preexisting physical topology

Can be computed by a mesh underprotocol, but…

Else Routing must infer its topology

Movement

natural and unescapable

Yet difficult to predict or detect


Peer selection

Potentially Large Peer Set

Highly Variable Capabilities

Selection Per Objective

Metrics (e.g. RSSI, ETX…)

L3 Reachability (::/0, …)

Constraints (Power …)


Constrained Objects

Smart object are usually

Small & Numerous

« sensor Dust »

Battery is critical

Deep Sleep

Limited memory

Small CPU

Savings are REQUIRED

Control plane

Data plane (Compression)


Fuzzy links

Neither transit nor P2P

More like a changing NBMA

a new paradigm for routing

Changing metrics

(tons of them!)

(but no classical cost!)

Inefficient flooding

Self interfering

QoS and CAC


Local Routing & Mobility

Stretch vs. Control

Optimize table sizes and updates

Optimized Routing Approach (ORA) vs

Least Overhead Routing Approach (LORA)

on-demand routes (reactive)

Forwarding and retries

Same vs. Different next hop

Validation of the Routing plane

Non Equal Cost multipath

Directed Acyclic Graphs (DAG) a MUST

Maybe also, Sibling routing

Objective Routing

Weighted Hop Count the wrong metric

Instances per constraints and metrics


Global Mobility

Pervasive Access

Satellite

3/4G coverage

802.11, 802.15.4

Always Reachable

at a same identifier

Preserving connections

Or not ? (CORE*, DTN**)

Fast roaming

Within technology (L2)

Between Technologies (L3)

* Constrained RESTful Environments

** Delay-Tolerant Networking


What’s missing

A radio abstraction

802.21, L2 triggers, OmniRAN

Roaming within and between technologies

A subnet model

NBMA, interference awareness

Federation via backbone / backhaul

Broadcast and look up optimization

Large scale

non-aggregatable

numbering and naming schemes

© 2010 Cisco and/or its affiliates. All rights reserved.BRKEWN-3012 50Unclassified

Why IPv6 ?


Why IP ?

Open Standards vs. proprietary

COTS* suppliers drive costs down but

Reliability, Availability and Security up

IP abstraction vs. per MAC/App

802.11, 802.15.4 (e), Sat, 3G, UWB

Keep L2 topology simple

To Infinity and Beyond… But End-to-End.

No intermediate gateway, tunnel, middle boxes & other trick

* Commercial, off-the-shelf


Things

Which IP version ?

The current Internet comprises several billion devices

Smart Objects will add tens of billions of additional devices

IPv6 is the only viable way forward

1~2 Billions

PCs & servers

Tens of

Billions

Smart Objects

Mobile

Fixed

2~4 Billions

Phones & carsIPv4 Unallocated pool to exhausted March 2011 !

RIRs pools to exhaust late 2011 and through 2012


Protocol Evolution

Little work on adapting IPv4 to radios

Rather adapt radios to IPv4 e.g. WIFI infrastructure mode

« Classical » IPv6

Large, Scoped and Stateful addresses

Neighbor Discovery, RAs (L3 beacons)

SLAAC (quick and scalable)

Anycast Addresses

IPv6 evolution meets Wireless:

NEMO (Mobile Routers) (Proxy) MIPv6

6LoWPAN ROLL/RPL

ISA100.11a ZigBee/IP


IPv6 addresses and headers

Stateful (states local and remote addresses)

Simple IPv6 Header

Extension Headers

Compressible

(6LoWPAN)


6LoWPAN Neighbor Discovery

Proactive Registration to the default Router (aka6LoWPAN Router, 6LR)

Default Router DADswith an Edge Router(aka 6LBR, B for Border)

ND proxy over a classicND backbone by Backbone Router(overloading 6LBR)


RPL: a 2-pass Routing Protocol forLow power and Lossy Networks (LLN)

1: DAG Information

Organize a routing topology

Distribute subnet information

Default route UP

2: Destination Advertisement

Advertise and install routes down

To prefixes, addresses and mcastgroup

Low control overhead

rapid convergence time

Or Energy conservation


NEMO & Global HAHA

Enables a subnet to change

its point of attachment to the

Internet

Packets to the mobile subnet

are forwarded by a Home

Agent over a dynamic tunnel

Nodes attached to MR are

unaware of the mobility

Global HAHA:

a global scalable model

See Also, LISP; HIP; PMIP…


IPv6 still lacks

NBMA / ML subnet

IPv6 only supports P2P and transit (ethernet)

By nature, a radio network is NBMA

L3 « VLAN »

So far only available with MPLS

Early attempts (MTR, RPL instances)

L4/5 hints

Flow Label given away to fwd plane

Microflows / compound flows

In WSN, a flow has multiple sources

Local and Global IP Mobility Unification

(eg MANEMO)


ISA100 and IETF

Introduced at ISA100, discussed at IETF

Split from the 6LoWPAN ND specWG decision (Hiroshima)

Added registration from RPL

No duplicate unique ID detectionAs discussed on the list, too complex


Translated in IETF terms

---+------------------------

| Internet/Plant Network

|

+-----+

| | Router / ALGateway

| |

+-----+

|

| Transit Link

+--------------------+------------------+

| | |

+-----+ +-----+ +-----+

| | Backbone | | Backbone | | Backbone

| | router | | router | | router

+-----+ +-----+ +-----+

o o o o o o

o o o o o o o o o o o o o o

o o o o o o o o o o o o o o o

o o o o o o o o o o

o o o o o o o

LLN LLN LLN


Initial time

---+------------------------


|

+-----+


| |

+-----+

|

| Transit Link

+--------------------+------------------+

| | |

+-----+ (root) +-----+ (6LBR) +-----+



+-----+ +-----+ +-----+

o o o o o o



o o o o o o o o o o

o o o o o o o

RPL LLN 6LoWPAN-only LLN LLN

A single subnet model

for the backbone and the

wireless sensor networks

Subnet

Route

In RIB

Subnet

Route

In RIB

Subnet

Route

In RIB

Subnet

Route

In RIB

Default

Route

In RIB


Registration (1ts step)

---+------------------------


|

+-----+


| |

+-----+

|

| Transit Link

+--------------------+------------------+

| | |

+-----+ (root) +-----+ (6LBR) +-----+



+-----+ +-----+ +-----+

o o o o o o



o o o o o o o o o o

o o o o o o o


DAO DAR

DAD DAD

Registration has:

• Lifetime

• Unique ID

• TID (SeqNum)


Registration (2nd step one second later)

---+------------------------


|

+-----+


| |

+-----+

|

| Transit Link

+--------------------+------------------+

| | |

+-----+ (root) +-----+ (6LBR) +-----+



+-----+ +-----+ +-----+

o o o o o o



o o o o o o o o o o

o o o o o o o


DAO

ackDAC

NA(O)

The BR maintains

a state and a route

to the WSN node

for the registration

lifetime

NA(O)

Host

Route

In RIB

Host

Route

In RIB


Duplication

---+------------------------


|

+-----+


| |

+-----+

|

| Transit Link

+--------------------+------------------+

| | |

+-----+ (root) +-----+ +-----+



+-----+ +-----+ +-----+

o o o o o o



o o o o o o o o o o

o o o o o o o


DAO

DAO

ack

(KO)

DAD NA

DAD option has:

• Unique ID

• TID (SeqNum)

Defend with NA if:

• Different UID

• Newer TID


Mobility

---+------------------------


|

+-----+


| |

+-----+

|

| Transit Link

+--------------------+------------------+

| | |

+-----+ (root) +-----+ +-----+



+-----+ +-----+ +-----+

o o o o o o



o o o o o o o o o o

o o o o o o o


DAO

DAO

ack

(OK)

DAD NA

DAD option has:

• Unique ID

• TID (SeqNum)

Defend with NA if:

• Different UID

• Newer TID


Resolution

---+------------------------


|

+-----+


| |

+-----+

|

| Transit Link

+--------------------+------------------+

| | |

+-----+ (root) +-----+ +-----+



+-----+ +-----+ +-----+

o o o o o o



o o o o o o o o o o

o o o o o o o


packet

NS

NA

NA option has:

• Unique ID

• TID (SeqNum)


Binding Tracking Option

Used to resolve conflicts

Need In ND: TID to detect movement

Need In RPL: Object Unique ID for DAD

+ DAO-ACK (DUPLICATE) flow

0 1 2 3

0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1

+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

| Type | Length | TID |

+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

| reserved |

+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

| |

+ Owner Unique Identifier +

| |

+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

Technology

Luxbg fringe