A lightweight process migration based computational offloading framework for mobile device augmentation / Abdullah

• Main Author: Abdullah, -
• Format: Thesis
• Published: 2017
• Subjects: QA75 Electronic computers. Computer science
• Online Access: http://studentsrepo.um.edu.my/7516/1/All.pdf; http://studentsrepo.um.edu.my/7516/5/abdullah.pdf; http://studentsrepo.um.edu.my/7516/

In recent years, the paradigm of mobile cloud computing has been introduced to extend capabilities of mobile devices, by taking advantage of high-speed wireless communications and high-performance cloud platforms to help gather, store and process data for the mobile devices. In this paradigm, the cloud-based mobile applications usually employ computational offloading for the augmentation of mobile device capabilities. Mobile device OS vendors are focused toward native mobile applications lifecycle to improve battery consumption and application execution performance. For example, Google has introduced Android Runtime Environment (ART) featuring Ahead of Time (AHOT) compilation to native instructions in place of Dalvik Virtual Machine (DVM) which consumes extra time and energy because of the Just in Time (JIT) compilation. However, current state-of-the-art offloading solutions do not consider AHOT compilations to native binaries in the ART environment. To address the issue in offloading ART-based mobile applications, we propose a lightweight computational offloading framework. The lightweightedness is measured as the overhead energy consumption and application execution time added up by the proposed framework. Further, we explain in details the design and implementation of the proposed prototype framework. The proposed framework requires infrastructural support from the remote computing platforms such as data centers or cloudlets to provide Offloading as a Service (OaaS) for a heterogeneous mobile cloud ecosystem. The proposed framework is evaluated using experimental testbed and validated using statistical modeling. Numerical results from the testbed revealed that the proposed framework saves almost 44% of the execution time and 84% of the energy consumption of the experimental application used.