Pervasive Mobile Applications

Post on 09-May-2022

5 views 0 download

Transcript of Pervasive Mobile Applications

Computational Offloading for Smartphones

Huber Flores huber@ut.ee

Tartu, Estonia, 2014

http://math.ut.ee/~huber/

Outline

• What is computational offloading?

– Cloudlets

– Code offloading

• Computational offloading in practice

– Android and Java reflection

• When computational offloading meets data analytics

Huber Flores 2

COMPUTATIONAL OFFLOADING

Huber Flores 3

What is computational offloading?

• Also known as Cyber-foraging

• Move a computational task from one place to another (aka offload)

• Why? – Constrained devices offload tasks to powerful

machines • Accelerate processing

• Save energy (battery life for mobiles)

• Adaptation of resource consumption to user’s experience

Huber Flores 4

Computational offloading

• Cloudlets

Huber Flores 5

Computational offloading

• MAUI

• ThinkAir

• CloneCloud

• EMCO, etc.

Huber Flores 6

Computational offloading

• Code offloading

Huber Flores 7

CODE OFFLOADING IN PRACTICE

Huber Flores 8

Demo: Chess offloading

• Minimax

Huber Flores 9

Exercise 1

• Java reflection

– https://gist.github.com/huberflores/9829019

Huber Flores 11

Exercise 2

• Transform a java method to a computational offloading format

– https://gist.github.com/huberflores/6e4e0894d95e16f1c25e

Huber Flores 12

Challenges and Issues

• Main problem: Computational offloading is unreliable.

– Negative/Positive

• Network communication

– Latency

– Data size, etc.

• Resource allocation

• Inference process

Huber Flores 13

When computational offloading meets data analytics

• Offloading in multiple dimensions

– What, when, how, where

• Quality of experience (QoE)

– Adaption of resource elasticity

Huber Flores 14

When computational offloading meets data analytics

Huber Flores 15

When computational offloading meets data analytics

• Pre-catching

– Mobile components

– Application states, etc.

Huber Flores 16

When computational offloading meets data analytics

Huber Flores 17

Beyond computational offloading

• More to come…

Huber Flores 18

QUESTIONS?

Huber Flores 19

References

• Suggested readings – Flores, Huber, et al. "Adaptive code offloading for mobile

cloud applications: Exploiting fuzzy sets and evidence-based learning." Proceeding of the fourth MobiSys workshop on Mobile cloud computing and services (MCS 2013). Taipei, Taiwan, 2013.

– Flores, Huber, et al. "Mobile code offloading: should it be a local decision or global inference?." Proceeding of the 11th annual international conference on Mobile systems, applications, and services (Mobisys 2013), Taipei, Taiwan, 2013.

Huber Flores 20

Home assignment 3 – Task 1

• Implement the following algorithms

– Quicksort

– Huffman

– Tower of Hanoi

• Transform into code offloading and try them via Java reflection

• Both versions of the algorithm can be executed

Home assignment 3 – Task 2

• Upload a picture or load one from the image gallery

• Picture is located into an small frame

• The frame can be drag and drop all over the screen

• The size of the frame is increased/decreased using gestures (pinch and spread)

• Once the frame reaches its maximum size, the faces in the picture are detected

Home assignment 3 – Task 2

Home assignment 3 – Task 2

Home assignment 3 – Task 2

Home assignment 3

• Deadline

– (09/11/2014 00:00)

– Task 1 – 10 pts

– Task 2 – 15 pts