fahad ali's slides for machine to-machine communication in rural conditions realizing...

15
Machine-to-machine communication in rural conditions: Realizing KasadakaNet By Fahad Ali Supervised by Victor de Boer

Upload: victor-de-boer

Post on 21-Jan-2018

337 views

Category:

Education


0 download

TRANSCRIPT

Machine-to-machine communication in rural conditions:

Realizing KasadakaNet

By Fahad AliSupervised by Victor de Boer

Contents

● Related works● Problem description ● Research question● Our solution: KasadakaNet● Evaluation● Discussion and conclusion

The Kasadaka Related works

● Enables information services forthe rural poor

● Rapid prototyping platform: allowseasy creation of applications for new use cases

M2M Communication Related works

● Transferring data between machines over some type of network● Over the internet, local Wifi/Bluetooth networks, cellular networks, etc.● M2M communication in rural conditions?

Problem description

● Lack of data sharing between devices● Pass-by approach:

sneakernet + local wifi networks● Case: Kasadaka platform

Research question

● RQ: "How can machine-to-machine communication and semantic data sharing be achieved using a pass-by communication method in such a way that it matches requirements from ICT4D use cases?"

System overviewKasadakaNet

● Sneakernet + local Wifi solution● Technical challenges:

○ Semantic data ○ Less human input

● Two components:○ A set of Kasadakas○ The Wifi-donkey

The Wifi-donkeyKasadakaNet

● Built using Open Source Software (PirateBox, ClioPatria, linux packages, custom scripts)

● Two main tasks:○ Host Wifi network and manage connected clients (hostapd, dnsmasq, works out of the

box)○ When a Kasadaka connects: send SPARQL query to Kasadaka, store RDF result (custom

scripts, configuring linux)

Experimental setup: measuring range Evaluation

Experimental setup: Pass-by experiment Evaluation

● Measured variables○ Success rate○ Result file size○ Request time

● Independent variables○ Travel speed○ Query size (# of triples)

Experimental setup: Scaling experimentEvaluation

● Load dataset of 475000 triples● Side-by-side data transfer● Measured variables:

○ Request time○ Result file size

● Independent variable:○ Query size in # of triples (7 categories: 30, 300, 1k, 5k, 10k, 50k, 100k)

Results: Pass-by experimentEvaluation

Results: Scaling experimentEvaluation

Discussion and conclusion

● Results show that the system works in the setting it was tested in.● Success rate depends mostly on time spent within range.● What about rural conditions and real scenarios?● And non-Kasadaka ICT4D implementations?

● A Sneakernet + local wifi solution is a viable approach in extending knowledge sharing systems with M2M capabilities:○ low cost hardware○ open source

Questions?