Wireless internet routing

Philippe Jacquet

Internet and networking

• Internet– User plurality connected to– Sources plurality

Wireless architectures

• Three main architectures– Point to point

– Base station

– Mesh

Signal processing complexity in ground wireless networks

• Diagram Capacity-Range

  

GSMUMTS

WavelanIEEE 802.11

Hiperlan1&2IEEE802.11a

UMTS pico-cell UMTS

micro-cell

Capacityin bit/s

distance in m

bluetooth

Mobile ad hoc networks

• Mesh and mobile

6

• DARPA’s proposal for LANdroids

2. Wireless network capabilities required

7

2. Which wireless networks?

MANET VANET wireless sensors

Wireless community network

Wireless networks

Mobile ad hoc networks– Mobility makes link failure a necessity

• Refresh period 1 second• Automatic self-healing

– Local neighborhood is local space• Unlimited neighborhood size

– Stadium network: N=10,000, with average degree 1,000

• BGP needs 1014 links exchange per refresh time• BGP fails on Wifi networks with 20 users at

walking speed.• Heavy density kills link state management.

Wireless internetMobile ad hoc routing

• Mobile Ad hoc NETworks– Two protocol classes:

• Proactive class• Reactive class

Proactive class

• Link state protocols– Full internet legacy– Topology compression– Periodic control traffic– Permanent routing tables– OLSR

Reactive class

• Distance vector based– Partial internet legacy– Path limited topology– On demand route discovery– Temporary routing tables– AODV

Optimized Link State Routing protocol

• Topology compression– Run light and fast, only on best routes

• Carry a subset of the local table • Stay on selected links

Topology compression graph(V,E)

• Optimized Link State Routing (OLSR) – A mobile network can be very dense (ex:

|V|=104,|E|=107)– Instead advertizing all links, a node advertizes only its

multipoint relays.– The multipoint relay set much smaller than neighbor

set |)(| EoEE r =→

)|(||| 2VOE =

Multipoint relay sets

Multipoint relay set of a terminal :Neighbor subset that covers the two hop neighbors

Goal: find the smallest posible multipoint relay set

Multipoint relay links

Multipoint relay set

• Local computation– Need to know two-hop neighborhood– Optimal is NP hard– Greedy algorithm is optimal to factor

• Good enough for performance

€

logV

Multipoint relay set

• Greedy selection algorithm

• On node AIntialize N2 N2(A), Er(A) .

While (N2) do

Find y in N(A) which maximizes |N2N(y)|

Add Ay to Er(A)

N2 N2- N(y)

Return Er(A)

Topology compression

• In Erdös-Rényi random graph models:

• In unit disk random graph model:

€

τ r =E rE

=Ologn

n

⎛

⎝ ⎜

⎞

⎠ ⎟

€

τ r =O n−

2

3 ⎛

⎝ ⎜

⎞

⎠ ⎟

BA

Topology compression

• The union of multipoint relay link sets is a remote-spanner– Nodes compute their routing tables with the

remote-spanner and their local table.

• Topology compression is lossless.– Optimal routes in tables also optimal in genuine

topology.

Route optimality proof

• By recursion: if D is at distance k from A in G(V,E) there exists a route of length k in G(V,E(A)+Er) from A to D– true for k=1: (A,D)E(A).– If true for k, then if D at distance k+1, let the chain F-H-D

with d(A,F)=k-1 and d(A,H)=k. Node F is a two hop neighbor of node D.

– Let M the multipoint relay of D which covers F.– M is at distance k from A, thus there exists a route L of

length k in G(V,E(A)+Er) from A to M.– (M,D) Er, thus L+(M,D) is in G(V,E(A)+Er) and is of length

k+1.A DF H

M

Control traffic reduction

• Topology reduction– Topology compression

• Dissemination reduction– Information about topology move only on

multipoint relay links

• Further dissemination reduction– Wireless flooding

• Total reduction

€

τ r

€

τ r

€

V

E

€

τ r2 V

E

Wireless flooding

• A single transmission per node– broadcast emission – Instead of a copy per link

Dissemination reduction

• link flooding (cost |E|) packet P node AP←reception();If !(P==null) then

if!(P received) theninsert P in received;B←last_emitter(P);for each C N(A) do

if(C≠B) then send(P,C);• broadcast flooding (cost |V|)

P←reception();If !(P==null) then

if!(P received) theninsert P in received;broadcast(P);

Multipoint relay flooding

• Only multipoint relays flood the information (reduction factor |Er|/|E|) P←reception();If !(P==null) thenif!(P received) then

insert P in received;B←last_emitter(P);

if(BA Er) then

broadcast(P);

Multipoint relay flooding

• Packet P dissemination from node S. Any arbitrary A– Closest emitter of P to A: distance k– Prove that k≤1

Multipoint relay flooding

A

FG

H

€

k

G’

€

k −1

€

k − 2

OLSR neighbor sensing

• Partial Link control on network layer

• Periodic broadcast of hello message– Contain list of heared neighbors– Two way check

MPR selection

• Collect the 2-way neighbor lists of neighbor nodes

• Run the selection algorithm.

• Advertize MPR set in hello