Robust Wireless Multicast using Network Coding

1
Robust Wireless Multicast using Network Coding

• Dawn Project Review,
• UCSC Sept 12, 06
• Mario Gerla
• Computer Science Dept, UCLA
• gerla_at_cs.ucla.edu www.cs.ucla.edu/NRL

2
Background Network Coding
Traditional multicast store and forward
3
Background Network Coding

• Network Codingstore-mix-forward

4
Network Coding wireless net
Store-mix-forward
a
a,b
a
a
a
b
a,b
b,a
optimal routingenergy per bit 5
network codingenergy per bit 4.5

• Wu et al. (2003) Wu, Chou, Kung (2004)
• Lun, Médard, Ho, Koetter (2004)

5
Random Network Coding
Sender
Every packet p carries e e1 e2 e3 encoding
vector prefix indicating how it is
constructed (e.g., coded packet p ?eixi where
xi is original packet)
x
y
z
A
ax ßy ?z
buffer
Random combination
Intermediate nodes randomly mix incoming packets
to generate outgoing packets
Destination
6
Problem Statement

• Multicast streaming in mobile wireless networks
  is non-trivial
• Streaming requires high reliability (but not
  100), low delay (but not 0)
• But network is unreliable, bandwidth-limited
• Major concern packet drops
• Lossy wireless channel (uncorrelated, random like
  errors)
• Route breakage due to mobility, congestion, etc
  (correlated errors)

7
Robust NC Multicast

• Most studies have evaluated NC M-cast in static
  networks no errors
• In tactical nets one must consider
• Random errors External interference/jamming
• Motion path breakage
• Target application
• Multicast (buffered) streaming
• Some loss tolerance
• Some delay tolerance (store playback at
  destination) - non interactive

8
Conventional vs NC Multicast

• Conventional Approaches
• Time diversity gt O/H, delay?
• Recovery scheme a la ARQ (Reliable Multicast)
• (End-to-end) Coding (FEC, MDC, )
• Multipath diversity (ODMRP, ) gt O/H?
• NC Approach
• Main ingredient Random network coding (by Médard
  et al., Chou et al.)
• Exploit every(?) diversity available
• Controlled-loss (near 100), bounded-delay
  (hundreds of ms)
• Suitable for buffered streaming
• Real time version (tens of ms delay bound)
  possible using progressive decoding

9
Network Coding in static wireless nets

• For cost efficiency
• Médard et al. Min-cost operation over coded
  Networks. IEEE T-IT
• Fragouli et al. A network coding approach to
  energy efficient broadcasting, INFOCOM 06
• Wu et al. Minimum-energy multicast in mobile ad
  hoc networks using network coding. IEEE TComm.
• For reliability
• Médard et al. On coding for reliable
  communication over packet networks.
• Others
• Ephremides et al. Joint scheduling and wireless
  network coding. In Proc. NETCOD 2005.

10
NC vs Conventional M-cast comparison

• Conventional Multicast ODMRP
• Mesh fabric Redundant paths
• Robust to motion and to errors

11
NC-Multicast evaluation

• Simulation study
• Scenarios with errors and motion
• Reported in IEEE Wireless Communication Magazine
  Oct. 2006 issue
• Performance bounds
• Static grid - corridor model
• Uniform, random errors
• Idealized MAC protocol (time slotting non
  interfering sets of hyperarcs)
• Linear programming optimal solutions
• Manually computed optimal solutions
• Reported in MILCOM 2006

12
Simulation experiments

• Settings
• QualNet
• 100 nodes on 1500 x 1500 m2
• 5 Kbytes/sec traffic (512B packet) - light load
• Single source multiple destinations
• Random Waypoint Mobility
• 20 receivers
• Metrics
• Good packet ratios num. of data packets received
  within deadline (1sec) vs. total num. of data
  packets generated
• Normalized packet O/H total no. of packets
  generated vs no. of data packet received
• Delay packet delivery time

13
ODMRP vs NC Reliability
Good Packet Ratio
14
ODMRP vs NC Efficiency
15
ODMRP vs NC Delay
16
ODMRP vs. NC Highway scenario
Randomly moving 200 nodes on 10kmx50m field. All
nodes are receivers.
17
Conventional Forwarding/Routing
forwarders
Source
Receiver
Select least number of nodes as forwarders to
form a path b/w a S-R pair and each forwarder
transmits each packet once
18
Problems with Conventional Routing
forwarders
Receiver
What if route breaks?
forwarders
What if random error occurs?
19
Network Code Forwarding
forwarders
Source
Receiver
Select most nodes in between a S-R pair as
forwarders and each forwarder transmits one
packet per generation once each node asks its
neighbors for more packets if it fails to get a
whole generation A node becomes a forwarder if
(hop count to Source hop count to Receiver) is
less than hop distance of S-R pair ?
20
Robustness of NC approach
Robust to mobility
Robust to random errors
21
Throughput Bounds

• Max NC-MCAST throughput in wireless networks?
• Previous simulation results based on light load.
  As load is increased, congestion leads to
  performance collapse
• Our approach evaluate max throughput
  analytically for a simple grid structure, the
  corridor

22
Linear Programming approach

• To calculate and compare maximum throughputs with
  and without NC, we use LP formulation
• Maximum multicast throughput LP models exist for
  wired networks
• We developed LP models for maximum throughput in
  unreliable wireless networks based on
• LP model developed for min-cost problems in
  unreliable wired network by Muriel et al.
• wireless medium contention constraints
• Also, we solve with LP for max throughput of
  conventional multicast (single tree and tree
  packing)
• LP solutions matched with manual solutions

23
Related Work Throughput Bound

• Previous works show the gap between NC and S/F
  for wired networks with no loss (e.g. log(n))
• For wireless networks
• Ephremides et al. Joint scheduling and wireless
  network coding. In Proc. NETCOD 2005.
• Wu et al. Network planning in wireless ad hoc
  networks a cross-layer. IEEE JSAC 2005.
• gt Both show throughput gain of NC calculated
  using link scheduling heuristics

24
Linear Programming Formulation

• maximize f

Wireless medium contention constraints
Wireless flow conservation constraints
25
Maximum Multicast Throughput Comparison NC vs
Conventional
CORRIDOR MODEL
Sender
Receivers
26
Network Coding Link schedule achieving
throughput of 2/3
A
B
C
D
A
B
C
D
B
A
(1)
(2)
(3)
(4)
(5)
(6)
E
F
G
H
H
G
F
E
F
E
D
C
A
B
C
D
AB
CD
(9)
(7)
(8)
(10)
(11)
(12)
27
Multicast with multiple embedded trees (no NC)
Link schedule achieves 2/5 throughput
B
A
A
B
A
B
A
(1)
(2)
(3)
(4)
(5)
D
C
C
D
C
D
C
B
(9)
(7)
(8)
(10)
(6)
28
An optimal Single Tree multicast schedule that
achieves 1/3
A
B
A
B
A
B
(1)
(2)
(3)
(4)
(5)
(6)
29
Future Work in Network Coding

• Implement NC - Mcast congestion control and ETE
  recovery above UDP
• If loss used as feedback, key problem is
  discrimination between random error and
  congestion
• TCP over Network Coded unicast
• Network Coding solutions for intermittent
  connectivity
• Models that include mobility

30
Vehicular Sensor Networks - Epidemic
Dissemination Models

• Car-Car or Car-Infostation communications using
  DSRC
• DSRC Dedicated Short Range Communication 802.11p
  IEEE Task group and derived from 802.11a

31
Vehicular Sensor Applications

• Environment
• Traffic congestion monitoring
• Urban pollution monitoring
• Civic and Homeland security
• Forensic accident or crime site investigations
• Terrorist tracking

32
Accident Scenario storage retrieval

• Private Cars
• Periodically collect images on the street (store
  data locally)
• Process the data and classify the event
• Create Meta-Data for event -- Summary (Type,
  Option, Location, Vehicle ID, )
• Post it on a distributed index
• The police access data from distributed storage

33
Epidemic Posting Harvesting

• Exploit mobility to create index and
  disseminate summaries
• Vehicles periodically broadcast summary of sensed
  data to their neighbors
• Data owner advertises only his own summaries
  to his neighbors
• Neighbors listen to advertisements and store them
  into their local storage
• A mobile agent (the police) harvests summaries
  from mobile nodes by actively querying mobile
  nodes
• Vehicles return all summaries collected so far

34
Epidemic Diffusion - Idea Mobility-Assist
Summary Diffusion
35
Epidemic Diffusion - Idea Mobility-Assist
Summary Diffusion
1) Periodically Relay (Broadcast) its
summary to Neighbors 2) Listen and store
others relayed summaries into ones storage
36
Epidemic Diffusion - Idea Mobility-Assist
Summary Harvesting

1. Agent (Police) harvestssummaries from its
   neighbors
2. Nodes return all the summariesthey have
   collected so far

37
Harvesting Analysis

• Metrics
• Fraction of harvested summaries F(t)
• Analysis assumption
• Discrete time analysis (time step ?t)
• N disseminating nodes
• Each node ni advertises a single summary si

38
Harvesting Analysis-Regular Nodes

• Expected number (a) of contacts in ?t
• ? density of disseminating nodes
• v average speed
• R communication range
• Incremental number of summaries harvested by a
  regular node ?Et Et - Et-1
• Prob. of meeting a not yet infected node is
  1-Et-1/N

39
Harvesting Analysis- Agent Node

• Agent harvesting summaries from its neighbors
  (total a nodes)
• A regular node has passively collected so far
  Et summaries
• Probability that agent can collect a specific
  summaryEt/N
• Specific summary collected from a neighbors with
  probability 1-(1-Et/N)?
• Let Et Expected number of summaries harvested
  by the agent

40
Harvesting Analysis - Harvesting Fraction

• Numerical analysis

Area 2400x2400m2Radio range 250m nodes
200Speed 10m/sk1 (one hop relaying)k2 (two
hop relaying)
41
Simulation

• Simulation Setup
• Implemented using NS-2
• 802.11a 11Mbps, 250m transmission range
• Network 2400m2400m
• Mobility Models
• Random waypoint (RWP)
• Urban map model
• Group mobility model
• Random Merge and split at intersections
• Westwood map

Westwood Area
42
Simulation

• Summary harvesting results with random waypoint
  mobility

43
Simulation

• Summary harvesting results with urban map mobility

44
Future Work

• Further investigate dependence of
  dissemination/harvesting from motion
• Enhance track models to reflect realistic (urban,
  open) scenarios
• Motion pattern characterization
• NCR (Neighborhood Change Rate)
• Fraction of traveling buddies, etc
• Data mining in large spatial-temporal databases
  on mobile platforms