Leveraging Content Connectivity and Location Awareness for...
Transcript of Leveraging Content Connectivity and Location Awareness for...
Leveraging Content Connectivity and Location Awareness for Adaptive Forwarding in NDN-
based Mobile Ad Hoc Networks
Muktadir Chowdhury1, Junaid Ahmed Khan2, Lan Wang1
1University of Memphis, 2Western Washington University
7th ACM Conference on Information-Centric Networking (ICN 2020)
2
Motivation
● Wireless networks are ubiquitous.● Mobile Ad-Hoc Networks are useful in many
scenarios.○ Vehicle-to-Vehicle Communication.○ Military Missions.○ Emergency Response.
● Data sharing enables vehicles to acquire information not in sensors’ range.• Avoid collision
• Know road information and change route accordingly.
Request
Requ
est
Request
NDN over MANET
● NDN has unique advantages over TCP/IP for wireless mobile communication.○ Name-based Forwarding [1]
○ Pervasive caching○ Network-Layer Security [2, 3]
● Develop a forwarding strategy for NDN-based MANET that can make smart forwarding decisions with minimal or no routing information.
3
Interest
Interest
Inte
rest
Data
Data
Data
[1] L. Wang, R. Wakikawa, R. Kuntz, R. Vuyyuru, L. Zhang, “Data Naming in Vehicle-to-Vehicle Communications”, IEEE INFOCOM Workshop on Emerging Design Choices in Name-Oriented Networking, 2012[2] M. Chowdhury, A. Gawande, L. Wang, "Secure Information Sharing among Autonomous Vehicles," ACM IoTDI, April 2017[3] M. Chowdhury, A. Gawande, L. Wang, "Anonymous Authentication and Pseudonym-Renewal for VANET in NDN," ACM ICN, poster Sept. 2017
Forwarding in NDN-based MANET
● VANET via Named Data Networking (VNDN) [1]○ Nodes farther from consumer will wait less time before forwarding.
● Navigo [2]○ Nodes closer to the Data location will be selected for forwarding.
● VNDN and NAVIGO are geo-location based routing.○ Need to resort to broadcast when location information is not available.
● STRIVE [3]○ Centrality Score → Interest satisfaction rate of a vehicle → A measure of
the connectivity of a vehicle. ○ Topology based: Vehicular Network is always changing à Frequently
update the table.○ Generalized score.
5
1. G. Grassi, D. Pesavento, G. Pau, R. Vuyyuru, R. Wakikawa, and L. Zhang."VANET via named data networking." IEEE INFOCOM WKSHPS, 2014.2. G. Grassi, D. Pesavento, G. Pau, and L. Zhang, "Navigo: Interest forwarding by geolocations in vehicular named data networking." IEEE WoWMoM, 20153. J. Khan and Y. Ghamri-Doudane. "Strive: Socially-aware three-tier routing in information-centric vehicular environment." IEEE GLOBECOM, 2016.
Content Connectivity and Location-Aware Forwarding (CCLF) Design
● Tableless timer-based forwarding.
○ Each node takes forwarding decision independently.
● Leverage Geo-Location in forwarding whenever available.
○ Not mandatory in Forwarding.
● Uses Centrality Score in Forwarding.
○ Interest satisfaction ratio.
○ Prefix-wise Centrality Score.
● Node-Density Aware Interest/ Data suppression
mechanism to encounter Broadcast storm.6
7
How CCLF Works - Overview
2
C
1
3
5
4
P
I=1, D=0S=0, L=0
I=1, D=0S=0, L=0
I=1, D=0S=0, L=0
I=9, D=8S=.72, L=1
I=9, D=9S=1, L=1
I=1, D=1S=1, L=1
I=1, D=1S=1, L=1
Packet Loss
6
I=4 D=2S=.50, L=1
I=0, D=0S=0, L=0
I=4, D=3S=.75, L=1
8
Prefix-based Content Connectivity Score
● Each node in the C-L tree has a name-prefix and it’s corresponding score and location.
● The score represents how successful a node has been in getting data under a name prefix.
● Score of a prefix j: 𝐶𝐶𝑆! ="!# ∑"∈$%&'(!) ""%!# ∑"∈$%&'(!) %"
● Periodically update Smoothed Score
C-L Tree
$𝐶𝐶𝑆&,( = 𝛼 𝐶𝐶𝑆&,( + (1 − 𝛼) $𝐶𝐶𝑆&,()*
Wait-time Calculation: Location Score
Distance of this node from Data location
Distance of Previous node from Data location
9
LS = 1 - "&+, (.,/)123("&+, .,/ ,"&+,(4,/))
● A node waits for FW-Delay timer before forwarding an Interest. The
timer is calculated using Location Score and Centrality Score.
p n1
d
n2
LS = 0.7
LS = 0.5
Data Location
10
Wait-time Calculation: Weight of a Node
● The weight of a node for a prefix:
● FW-Delay timer of the node for the prefix :
Smoothed Centrality Score for the prefix
Location Score for the prefix (0 if Location not available)
At the beginning, when a node doesn’t have Location Information or Centrality Score
𝑇 is an upper bound on 𝑡 . The forwarding timer is set to a random value uniformly distributed between 0.5𝑡 𝜇s and 1.5𝑡 𝜇s
11
Density Aware Packet Suppression
● Receives same Interest before timer expires à Calculates suppression probability
𝐾 = 𝑆𝑢𝑝𝑝𝑟𝑒𝑠𝑠𝑖𝑜𝑛 𝐶𝑜𝑛𝑠𝑡𝑎𝑛𝑡, 𝑛 = 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑁𝑒𝑖𝑔ℎ𝑏𝑜𝑟𝑠
● Higher the number of neighbors, higher probability of suppression.
● We have run experiments to find a suitable value for K in suburban and urban environment.
𝑝 = min(𝐾 ∗ 𝑛, 1)F1
F2
1
1
2
2
4
3
F2 will have a higher probability to cancel forwarding and suppress a received packet.
Ad-Hoc Link Adaptation Layer (ALAL)
● Enhance reliability of packet
forwarding in Mobile Ad-hoc
Network.
12
● Roles of ALAL
1. Keep track of number of the neighbors in Tx range.
2. Queue outgoing packet when no neighbors around.
3. Get location information from GPS and add Location header.
Obtaining Geo-location and Prefix Information
● Add Location Header in NDNLP[1] packet.
○ Previous node’s location
○ Data location
● Interest can carry the Location header.
● Data can carry Data-Location field and Prefix Announcement header.
13
● C-L tree will be populated by the Prefix Announcement and Location in the Data packet
[1] Shi, Junxiao, and Beichuan Zhang. "NDNLP: A link protocol for NDN." NDN, NDN Technical Report NDN-0006 (2012).
How does CCLF address existing schemes’ limitations?
● Topology-based proactive schemes:
○ High overhead ß CCLF has no routing messages
● Topology-based reactive schemes:
○ High delay ß CCLF uses name-prefix based connectivity
score to avoid initial learning delay.
● Pure geo-location based schemes:
○ Local optimum and broadcast ß CCLF uses connectivity
score to get around local optimum and avoid broadcast
storm.14
15
Implementation and Evaluation
● Forwarding Strategy is implemented in ndn-cxx library and NFD.○ Added NDNLP headers in ndn-cxx.○ Forwarding Strategy in NFD.○ AdHoc Link Adaptation Layer in Transport Service
● Simulation using ndnSIM.○ Traffic Trace and Map: SUMO[1]
● Metric○ Interest Satisfaction Ratio○ Data Fetching Delay○ Message Overhead
Forwarding Strategy
Link Service
NDNPacket
Transport Service(AdHoc Link
Adaptation Layer)
Net Device
NDN packet + Current Location
Transport packet
Ns3 packetNs3 packet
Transport packet+ Current Location
ndnS
IM N
FD
Mod
ule
[1] Krajzewicz, Daniel, et al. "SUMO (Simulation of Urban MObility)-an open-source traffic simulation." Proceedings of the 4th middle East Symposium on Simulation and Modelling (MESM20002). 2002.
16
Simulation Parameters (Vehicular Topology)
Communication Parameter Setting Traffic Parameter Setting
Parameters Value
Propagation Loss Model Range Propagation
WiFi Transmission Range 100 m
Wifi Standard 802.11p
Physical Mode OfdmRate6MbpsBW10MHz
Propagation Delay Model Constant Speed
Data Packet Payload Size 1200 B
Parameters Value
Road Topology 3x3 Grid
Area 300x300 m2
Speed 7-13 m/s
Value 1 Value 2 Value 3
Value 4 Value 5 Value 6
Value 7 Value 8 Value 9
(0,0) (0,300)
(300,0) (300,300)
Intersectionof tworoads
100
m
Speed: 7-13 m/s
17
Determining the value of K
● Suppression Constant, K = 0.04.
● Performance of CCLF for various value of K in Grid topology.
17
Traffic Condition
No. of Vehicles/Mile
/ Lane
Median No. of Neighbors (2
Lane)
Median No. of Neighbors
(Grid)
Duplicate Suppression
Probability (2 Lane)
Duplicate Suppression
Probability (Grid)
Sparse 1-12 1-4 1-11 4%-16% 4%-44%
Medium 13-30 5-10 12-26 20%-40% 48%-100%
Dense 31-40 10-15 26-34 40%-60% 100%
20 30 40 50 60 70 80
No. of Vehicles
0.0
0.2
0.4
0.6
0.8
1.0
Nor
mal
ized
Pro
toco
lO
verh
ead
K=.04
K=.08
K=.12
Flooding
20 30 40 50 60 70 80Number of Vehicles
0.0
0.2
0.4
0.6
0.8
1.0
Sat
isfa
ctio
nRat
io
K=.04
K=.08
K=.12
Flooding
20 30 40 50 60 70 80
No. of Vehicles
0
1
2
3
4
5
6
Del
ayStr
etch
K=.12 (Median)
K=.08 (Median)
K=.04 (Median)
K=.12 (95th)
K=.08 (95th)
K=.04 (95th)
18
Results: CCLF (With Geo Location)
20 30 40 50 60 70 80
No. of Vehicles
0.0
0.2
0.4
0.6
0.8
1.0
Nor
mal
ized
Pro
toco
lO
verh
ead
CCLF
VNDN
NAVIGO
Flooding
20 30 40 50 60 70 80Number of Vehicles
0.0
0.2
0.4
0.6
0.8
1.0
Sat
isfa
ctio
nRat
io
CCLF
VNDN
NAVIGO
Flooding
20 30 40 50 60 70 80
No. of Vehicles
0
1
2
3
4
5
6
7
Del
ayStr
etch
CCLF (Median)
NAVIGO (Median)
VNDN (Median)
CCLF (95th)
NAVIGO (95th)
VNDN (95th)
● One consumer and One producer.● Consumer sending one interest/second.● Comparing CCLF with other geo-
location-based strategies: VNDN and NAVIGO.
19
20 30 40 50 60 70 80Number of Vehicles
0.0
0.2
0.4
0.6
0.8
1.0
Sat
isfa
ctio
nRat
io
CCLF
STRIVE
Flooding
20 30 40 50 60 70 80
No. of Vehicles
0.0
0.2
0.4
0.6
0.8
1.0
Nor
mal
ized
Pro
toco
lO
verh
ead
CCLF
STRIVE
Flooding
20 30 40 50 60 70 80
No. of Vehicles
0
2
4
6
8
10
12
Del
ayStr
etch CCLF (Median)
STRIVE (Median)
CCLF (95th)
STRIVE (95th)
Results: CCLF (W/O Geo Location)
● One consumer and One producer.● Consumer sending one interest/second.● Comparing CCLF with a centrality score-
based strategy : STRIVE.
20
Conclusion
● CCLF incurs lower protocol overhead ß Less Flooding + Density-aware packet suppression + Packet queuing at ALAL.
● CCLF incurs lower delay because Name-prefix based connectivity helps:
○ To avoid initial learning delay.
○ To find better connected path to data producer.
● Future Works
○ Study different ways to calculate forwarding timer and suppression probability.
○ Offer guidelines on setting CCLF parameters.