RMCC: A RESTful Mobile Cloud Computing Framework for Exploiting Adjacent Service-based Mobile Cloudlets
1. RMCC: A RESTful Mobile Cloud Computing Framework for
Exploiting Adjacent Service-based Mobile Cloudlets
Saeid Abolfazli (PhD)
Center for Mobile Cloud Computing Research
University of Malaya
Dec 2014
Presented in IEEE CloudCom’14 Conference, Singapore
15-19 December 2014
2. Motivation
• Trend: Mobile Everywhere
• However: Intrinsic Resource Poverty
=
Constraint CPUShort Battery Life Small Storage
3. State-of-the-art: Mobile Cloud Computing
• Leverage cloud-based resources
• Augment mobile devices
• Perform resource-intensive task remotely
• Major issues with tradition augmentation frameworks:
1. WAN latency
2. Partitioning overhead
3. Portability
4. RMCC main idea and use cases
• Use ASMobiC: Adjacent (one-hop) service-based mobile cloudlets as computing server
Resource sharing Incentive:
- Financial benefits (at least electricity bill)
- Reputation
- Reputation-based mutual benefits
Feasible Use cases
- Distributed analysis of sensitive/confidential/enterprise data
- Online real-time OCR in hospital
- E-learning in group
- On-campus scientific computing
- On-road navigation
- Real-time computing for smart city
5. RMCC Design Considerations & Significance
• Service-oriented architecture (loose coupling)
• Separation of responsibilities (simple and convenient)
• No code offloading (less data transfer)
• REST web services (less overhead, stateless)
• Arbitrated by MNO (mobile network operators)
• Centralized/decentralized mode (flexible security)
• Asynchronous
• Internet-free
• Green Computing
6. RMCC Architecture
• Main components:
Mobile Service Consumer
Mobile Service Provider
Trusted Service Governor
7. Evaluation
Methods:
1- Mathematical Modeling (Statistical Modeling)
2- Benchmarking
Evaluation Metrics and tools:
1- Application Execution Time (ms) - > Auto-logging
2- Mobile Consumed Energy (mJ) -> Power Tutor 1.4
Entity Specification
Mobile Service Consumer HTC Nexus One, Android-based
Wireless Access Point Cisco Linksys WRT 54G
Mobile Service Provider 1 Samsung Galaxy S2
Mobile Service Provider 2 Dell Laptop XPS 14x
Mobile Service Provider 3 Acer Laptop
Centralized Server Dell OptiPlex 990
Database SQL Server
Number of Workload 30
8. Statistical modelling
Via Linear Regression Model
• Generate: Independent Replication Method
• Train regression model using measured dataset to derive
regression equation.
• Derive model of time and energy via algorithm complexity
(Big-O) and regression equation.
• Validate using split-sample approach
• Generate time and energy data.
• Synthesize the results