HINT: from Network Characterization
to Opportunistic Applications
ACM CHANTS 2016, New-York
Gwilherm Baudic, Antoine Auger, Victor Ramiro, Emmanuel Lochin
Introduction
✓ More and more connected devices
✓ Opportunistic networks are a promising
alternative to infrastructure-based networks
➡
Opportunistic applications
= user applications
running over opportunistic networks
OppChat
: a simple
messaging application
Opportunistic applications behavior
is highly affected by network
p e r f o r m a n c e s ( Q o E , Q o S ,
application footprint)
Developers want a convenient tool
to see if their application works!
OppChat
: a simple
messaging application
Opportunistic applications behavior
is highly affected by network
p e r f o r m a n c e s ( Q o E , Q o S ,
application footprint)
Developers want a convenient tool
to see if their application works!
➡ There is a gap between network
characterization and application
Opportunistic network
characterization
Developers of opportunistic applications should have
some knowledge about network characterization
➡ Impor tant to understand and improve
application behaviour
Opportunistic network
characterization
Developers of opportunistic applications should have
some knowledge about network characterization
➡ Impor tant to understand and improve
application behaviour
Analytical modeling Simulators Traces collection Emulators and testbedsH
int
I
s
N
ot a
T
estbed
We propose the HINT emulator to bridge the gap between network characterization and application development
Main requirements: Real-time, Contact-oriented emulation, Tuning, Monitoring, Application transparency, Repeatability, Availability
Emulated opportunistic network
The HINT emulator
HINT relies on the definition of user scenarios
➡ A scenario is programmatically defined writing a
Python script that implements the class
ScenarioThe HINT emulator
HINT relies on the definition of user scenarios
➡ A scenario is programmatically defined writing a
Python script that implements the class
ScenarioA user
Scenarioshould *at least* define:
•
List of real and virtual nodes
•List of node pairs
•
Pairwise Contact frequency (Analytical models or
Traces)
The HINT emulator
Core Emulator Message Broker Monitoring & Tuning Database Real world HINT emulator App ULL App ULL App ULLHINT evaluation:
Application development
Message metadata: w e s e e h o w t h e message traverses several opportunistic nodesHINT evaluation:
Application development
1. Include the User Link
Layer (ULL) package
HINT evaluation:
Application development
1. Include the User Link
Layer (ULL) package
2. Start the
HINT evaluation:
Application development
1. Include the User Link
Layer (ULL) package
2. Start the
UserLinkLayerService
3. Use APIs to send /
receive opportunistic
messages
Monitoring and tuning:
✓ Network view
✓ Pair view
✓ Node view
Monitoring and tuning:
✓ Network view
✓ Pair view
✓ Node view
A demo of HINT is scheduled later atHINT evaluation:
Core scalability
Trace Average Maximum
Rollernet 26 146
MIT 180 days 0.008 34
Infocom 2005 0.22 26
Average and maximum number of events per second for real traces
The Core Emulator of HINT can schedule around 300 events/s with no more than 1s
threshold delay Currently using Python basic scheduler (proof of concept) 300 0.00 0.25 0.50 0.75 1.00 10 1000 Missed e vents r atio T=1 T=0.1 T=0.01
Conclusions
✓ HINT is a lightweight hybrid emulation system to
help developers of opportunistic applications
Conclusions
✓ HINT is a lightweight hybrid emulation system to
help developers of opportunistic applications
✓ Application transparency, network monitoring/tuning
Conclusions
✓ HINT is a lightweight hybrid emulation system to
help developers of opportunistic applications
✓ Application transparency, network monitoring/tuning
➡ Future work: extended evaluation
➡ New routing protocols, buffer management policies, core scheduler implementation (APS)
Conclusions
✓ HINT is a lightweight hybrid emulation system to
help developers of opportunistic applications
✓ Application transparency, network monitoring/tuning
➡ Future work: extended evaluation
➡ New routing protocols, buffer management policies, core scheduler implementation (APS)
Antoine Auger
antoine.auger@isae.fr
Thank you for your attention
This research was supported in part by the French Ministry of Defense