Date post: | 21-Jan-2018 |
Category: |
Education |
Upload: | huber-flores |
View: | 106 times |
Download: | 0 times |
Crowdsensing the opportunistic context of mobile devices
Helsinki, Finland.
EVIDENCE-AWARE MOBILE COMPUTATIONAL
OFFLOADING Huber Flores, Pan Hui, Petteri Nurmi, Eemil Lagerspetz, Sasu Tarkoma, Jukka
Manner, Vassilis Kostakos, Yong Li and Xiang Su
Outline
• Background
– Mobile code offloading
• Motivation
• Problem statement
• Evidence-aware mobile computational offloading
• Implications for the edge
• Conclusions
Helsinki, Finland.
2
Background
• Opportunistic augmentation of resources
Helsinki, Finland.
[IEEE Communications] Flores, H., Hui, P., Tarkoma, S., Li, Y., Srirama, S., & Buyya, R. (2015). Mobile code offloading: from concept to practice and beyond. IEEE Communications Magazine, 53(3), 80-88.
3
Motivation
• The offloading outcome is diverse due to many parameters – Latency – Code profiling – Device workload – Server processing the task – Type of device – Etc.
• Initial idea…. – Can we do better?
• …. Tuning parameters
Helsinki, Finland.
4
Problem?
• It is not that easy
– For a single device
• Is it possible to crowdsource the problem?
Helsinki, Finland.
5
Crowdsensing characterization
Helsinki, Finland.
6
Crowdsensing characterization
Helsinki, Finland.
[IEEE Communications] Flores, H., Hui, P., Tarkoma, S., Li, Y., Srirama, S., & Buyya, R. (2015). Mobile code offloading: from concept to practice and beyond. IEEE Communications Magazine, 53(3), 80-88.
7
Crowdsensing characterization
Helsinki, Finland.
8
Crowdsensing characterization
Helsinki, Finland.
9
Crowdsensing characterization
Helsinki, Finland.
10
Crowdsensing characterization
Helsinki, Finland.
[ICDCS] Flores, Huber, et al. “Modeling Mobile Code Acceleration in the Cloud" , Proceeding of ICDCS 2017, Atlanta, USA, June 5-8, 2017.
11
Crowdsensing support
• LAPSI
Helsinki, Finland.
12
EMCO framework
Helsinki, Finland.
13
Crowd evaluation
Helsinki, Finland.
14
Crowd evaluation
Helsinki, Finland.
15
Off-the-shelf applications
Helsinki, Finland.
16
Implications
• Combine cloud/edge provisioning – Dynamic allocation of resources
• Self-organizing systems (end points at the edge) – Offloading
– Sensing
– Networking
– Storage
– And so on
Helsinki, Finland.
17
Implications
• Edge infrastructure
Helsinki, Finland.
18
Implications
• Edge infrastructure
Helsinki, Finland.
Task
19
Summary
• Context characterization is really important, but it is not a task a single device can perform
• We demonstrate how context adaptation improves offloading
• A methodology for context reconstruction from passive data (datasets)
Helsinki, Finland.
20
QUESTIONS
Helsinki, Finland.
21