Content Distribution in VANETs using Network Coding: The Effect of

Content Distribution in VANETs using Network Coding:
Evaluation of the Generation Selection
Algorithms
Alexander Afanasyev
Tutors: Seung-Hoon Lee, Uichin Lee
March 2, 2009
Content Distribution in
Vehicular Ad-Hoc Networks (VANETs)

Applications
◦ Software updates and patches (e.g., navigation map, games)
◦ Multimedia data downloads (e.g., videos, news, etc.)
Content Distribution Challenges
◦ High mobility (i.e., highly dynamic networks)
◦ Error-prone channel (due to obstacles, multi-path fading, etc.)
CarTorrent: BitTorrent-like Cooperative
Content Distribution in VANETs
A file is divided into pieces
Web
Server
Exchange pieces via Vehicle-to-Vehicle Communications
Download a file (piece by piece)
Not useful!

Problem: Peer & Piece selection
 coupon collection problem
Cannot complete download!
Using Network Coding: CodeTorrent
A file is divided into pieces
Web
Server
1 more?
Any linearly independent coded packet is helpful

Network Coding Problem
Processing Overhead
Single Generation
Overhead
5/10/50 Generations

Delay without O/H
◦ Small # of generations is a better choice
◦ Larger # of generations  more severe coupon collection problem
Mitigating Coding Overheads

Solution: divide a file into small generations
◦ Problem: too many generation causes a coupon
collection problem
◦ Conflicting goals: maximizing benefits of NC vs.
minimizing coding O/H
50MB
Mitigating Coding Overheads

Solution: divide a file into small generations
◦ Problem: too many generation causes a coupon
collection problem
◦ Conflicting goals: maximizing benefits of NC vs.
minimizing coding O/H
10MB x 5
1
4
What is optimal strategy for
generation downloading?
Global:
(neighbor status)
Gen1
Gen2
Gen3
Local:
(my status)
Gen1
Gen2
Gen3
Request to ??


Checking neighbor rank improve chances of linearly
independent block, but
◦ Low-rank cars can also have valuable blocks
Back to the BitTorrent problem of piece/generation selection
◦ Local status based decision (i.e., the least/the most
downloaded generation, sequential order)?
◦ Neighbor status based decision?
◦ Random?
Generation Selection Strategies
Virtual “Global”
Completeness Vector
Global Min: Gen 4
Global Max: Gen 3
Random: Random
Sequential: Gen 1
Simulation Setup

Communications
◦ 802.11b; 11Mbps + Two-ray ground propagation

Mobility
◦ Random Waypoint model w/ speed range of [0,20] m/s
◦ Westwood area map: 2400m*2400m

Nodes
◦ 3 APs: file sources
◦ 200 nodes/40% interest level:
 80 nodes are downloading a file

Download parameters
◦ 50 megabyte file
◦ 10 generations
Westwood area map
Downloading all generations in parallel
Generation Progress
Global Min
Neighbor Min
Local Min
Random
Downloading all generations in parallel
Overall Progress
Neighbor-aware
strategy improves
at the beginning of
downloading
Local-aware and
random strategies
has smaller tail
* confidence interval is calculated with probability 95% using 8 simulations
Downloading all generations in parallel
Finishing times histogram
Conclusions:
 Network-aware strategy has long tail of finishing times
 Local and random strategies behave almost as good as global status-aware
Downloading generations (semi-)sequentially
Generation Progress
Global Max
Neighbor Max
Local Max
Sequential
Downloading generations (semi-)sequentially
Overall Progress
!!! Neighbor-aware
strategy outperforms
local and global one
* confidence interval is calculated with probability 95% using 8 simulations
Downloading generations (semi-)sequentially
Finishing times histogram
Conclusions:
 Network-aware strategy outperforms other strategies
 Average finishing time for global/local max strategies 1.5 times worse than
neighborhood status aware policy
Parallel vs Sequential Downloading
Overall progress of the best strategies
Conclusions:
 Neighbor-aware generation choosing considerably improves chances for
helpful block (linearly independent) at the beginning
 Local or random strategy improves download finishing time
Parallel vs Sequential Downloading
Finishing times histogram
Conclusions:
 Neighbor-aware strategies have on average 20% worse finishing times than
local max strategy
Interesting Facts


Checking generation rank of
the available generation
greatly improves performance
for neighbor status aware
strategies
Integer vector gossiping
decrease overall download
performance
Rank
checking
No rank
check
Bool
vector
Int
vector
Conclusion





Generation selection strategy in multi-generation
CodeTorrent downloads have big impact on the overall
download performance
Local status aware strategies (local-min, random) have the
best finishing performance
Neighbor status aware strategies have the best start-up
performance
It is important to check rank for neighbor status aware
strategies
Future work
◦ Investigate performance of combined strategies
◦ Check performance using different node mobility models