Integration of ApplicationLayer Scheduling and Routing in DelayTolerant MANETs

1
Integration of Application-Layer Scheduling and
Routing in Delay-Tolerant MANETs

• José Brustoloni, Sherif Khattab, Christopher
  Santamaria, Brian Smyth, and Daniel Mossé
• CS_at_PITT

2
Mobile Ad-Hoc Networks (MANETs)
Cell phone
3
Mobile Ad-Hoc Networks (MANETs)
4
Communication of First-Responders

• Connecting handheld devices used to exchange
  data, images, video, voice, work orders, etc.

5
Need for MANETs

• MANETs are
• needed if traditional
• communication
• infrastructure is
• damaged

6
Network Partitioning

• MANETs may get partitioned
• Field characteristics
• Noisy environments

Packethop.com
7
Delay-Tolerant Networking

• DTN research deals with routing of messages
  across network partitions
• Our work proposes a new approach to DTN routing

8
Example Scenario
Main Partition
Subordinate Partition
Leader
EOC
9
Example Scenario
Main Partition
Subordinate Partition
Work Order
Leader
EOC
10
Example Scenario
Main Partition
Subordinate Partition
Work Order
Courier
Leader
EOC
11
Work Order Model
Execution Time
Deadline
Pre-emption
Deadline Miss
Time
12
Work Order Model
Each task associated with a location
Deadline
13
Courier Selection Problem
Main Partition
Subordinate Partition
?
?
?
Work Order
Courier
Leader
EOC
14
Metrics

• Percentage of missed deadlines
• Average traveled distance per node

Number of deadlines missed
Total number of work orders
15
State-of-the-art

• Dedicated mobile elements
• Message Ferries _at_GeorgiaTech handle only
  message delivery
• Minimize average delay
• Trajectory modification of mobile users
• _at_Dartmouth Mobicom00
• Minimize detour distance

16
Our Hypothesis

• We can achieve better trade-off between missed
  deadlines and traveled distance if
  application-layer demand is taken into
  consideration in courier selection

17
Highest-Slack Courier Selection
Main Partition
Subordinate Partition
Maximum Leeway
18
Compared Schemes

• Closest
• Select courier closest to work order destination
• Dedicated
• Set of nodes dedicated for message delivery
  (dont execute any work)
• Random

19
Common Assumption

• Leader aware of current position of
  main-partition workers
• GPS-enabled devices
• Landmarks

20
Simulation Parameters

• Rate of work orders (load)
• main and sub-ordinate
• default 60
• Distance between partitions
• default 1200m
• Number of dedicated couriers
• default 1
• Speed of dedicated couriers
• default 5 m/s (18 km/h)

21
Distance between Partitions
100
Dedicated
Random
Closest
Highest-Slack
80
60
Deadlines Missed
40
20
1200m
200m
400m
800m
22
Subordinate-partition Load
100
Random
Closest
Dedicated
Highest-Slack
80
60
Deadlines Missed
40
20
20
40
60
80
23
Why?
Main Partition
Subordinate Partition
24
Subordinate-partition Load
16km
12km
Random
Traveled Distance Per Node
8km
Closest
Highest-Slack
4km
Dedicated
20
40
60
80
25
Main-partition Load
100
Dedicated
Random
Closest
Highest-Slack
80
60
Deadlines Missed
40
20
20
40
60
80
26
Dedicated-courier Speed
100
Random
Closest
Dedicated
Highest-Slack
80
60
Deadlines Missed
40
20
54 km/h
18 km/h
25 km/h
27
Number of Dedicated Couriers
100
Random
Closest
Dedicated
Highest-Slack
80
60
Deadlines Missed
40
20
5
1
10
15
20
28
Conclusions and Future Work

• Courier Scheduling in partitioned ad-hoc networks
• Integrated application- and network-layer
  scheduling
• More realistic
• models of work orders
• metrics (e.g., rate of casualties)
• frequency and structure of network partitions
• Comparison with other schemes
• communication bridges

29
Questions ?
30
Work Order Parameters

• Average Deadline 440 sec
• Execution Time 0.5 Deadline
• Enough to run back and forth across a 500m
  partition and still meet deadline

31
Simulation Time
Warm-up
Statistics Gathering
Cool-down
10000 ( 2.5 Hrs)
10
1000
9500
Time (Seconds)
32
When to return home?
Main Partition
Subordinate Partition
33
(No Transcript)