Project Presentation

Download Report

Transcript Project Presentation

Net-Centric Software and
Systems I/UCRC
A Framework for
QoS and Power Management for
Mobile Devices in Service Clouds
Project Lead: I-Ling Yen, Farokh Bastani, Krishna Kavi
Date: October 21, 2010
Copyright © 2010 NSF Net-Centric I/UCRC. All rights reserved.
2010/Current Project Overview
A Framework for QoS and Power Management for
Mobile Devices in Service Clouds
Project Scope:
Project Schedule:
Task 6: Evaluation
Image
capturing
Audio
capturing
…
Personal
Info DB
Facial
Image DB
Translation
Service
Service Cloud
Personal
Info DB
Translation
Service
Local Info
Service
Video
DB
Facial
Image DB
Execute the services in
Cloud or mobile device?
QPM framework Save power & satisfy QoS req.
Tasks:
1. Build the experimental environment
2. Develop the prediction algorithms to predict
QoS and power behavior for each service
and for a service chain
3. Develop the execution decision algorithms
4. Develop the service migration infrastructure
5. Develop the service allocation decision
algorithms
6. Validate the framework design
Tasks 2,3: Simple data
collection + coordinated
prediction & decision
Task 1
Tasks 1-3,6: Enhanced
data collection,
prediction & decision
Task 6: New
Evaluation
A M J J A S O N D J F M A
10
11
Deliverables:
• Experimental results showing the benefit of
using service cloud in saving power for
mobile devices
• Design of power optimization algorithms
Success Criteria:
• This project will demonstrate a significant
improvement in reducing power consumption
on mobile devices by delegating tasks to
service cloud
7/12/2016
Page 2
Significant Finding/Accomplishment!
Complete
Partially Complete
2009 Project Results
TASK
1. Build the experimental
environment
2. Develop the prediction
algorithms to predict QoS and
power behavior for each
service and for a service chain
3. Develop the execution decision
algorithms
Not Started
STAT
PROGRESS and ACCOMPLISHMENT

Set up laptop and PC as the mobile device and
the service cloud. Setup PowerTop for mobile
device power measurement.

Collected data and used them as historical
information for prediction. Developing neural
network to make QoS and power predictions
for unexplored configurations.

Completed a decision algorithm for the mobile
device to determine whether to execute the
services in a task on the mobile device or in the
service cloud.

Established 3 scenarios, developed the
involved services, used them to validate the
framework design and obtained some results.
6. Validate the framework design
This research leverages service clouds for significantly reduced power
consumption & latency on mobile devices
7/12/2016
Page 3
Our Solution
•
Use service cloud to help power management in mobile devices
• E.g., computation intensive services can be delegated to cloud
• E.g., communication intensive services can be migrated to MD
•
Decision process
• Service execution platform selection decision (SEPSD)
• Service allocation decision module (SADM)
• Service migration infrastructure (SMI) in the cloud
• Offline analysis in the service cloud
to determine the best QoS and
power management parameters QPMMD
 Derive parameterized rules
SPESD
SADM
SPESD
QPM
SADM
SMI
• Mobile device makes on-the-fly
decisions based on the rules
QoS-PM
Service Profile
QoS-PM
CAM-AM
User Profile
Power Manager
Service
Profile
User
Profile
Resource
Profile
7/12/2016
Page 4
Major Accomplishments, Discoveries
and Surprises
1. Evaluation results (power saving by QPM)
• Template holder, new results will be added for presentation
2. Developed a QPM pattern to be submitted to NCOIC
• Include a complete design of the system with major components
that can be implemented using different technologies
7/12/2016
Page 5
New Problems
•
•
•
Based on the experimental results, Improve the QoS and power
prediction accuracy by using better prediction models
•
Consider more parameters that may affect QoS
•
Train the prediction model with service profiles collected preliminary
experimental studies
Optimize SEPSDM decision process to reduce its power & latency
•
For each service, make pre-analysis for mobile device and user specific
predictions before downloading the service
•
For frequently used service chain, pre-compute the tradeoff in QoS and
power and maintain them in a table to facilitate quick run time search of
best decisions
Develop the data migration decision techniques in SADM
•
Currently we use static decision on which data migration policy to use for
each specific application (provided in the service profile)
•
Will consider dynamic approach and implement prototypes
7/12/2016
Page 6