A Task Offloading Framework for Energy Saving on Mobile Devices using Cloud Computing
Year: 2015
Type: dissertation
Abstract: Over the last decade, mobile devices have become popular among people, and their number is ever growing because of the computing functionality they offer beyond primary voice communication. \nHowever, mobile devices are unable to accommodate most of the computing demand as long as they suffer the limited energy supply caused by the capacity of their small battery to store only a relatively small amount of energy. \nThe literature describes several specialist techniques proposed in academia and industry that save the mobile device energy and solve this problem to some extent but not satisfactorily. \nTask offloading from mobile devices to cloud computing is a promising technique for tackling the problem especially with the emergence of high-speed wireless networks and the ubiquitous resources from the cloud computing. \nSince task offloading is in its nascent age, it lacks evaluation and development in-depth studies. \n \nIn this dissertation, we proposed an offloading framework to make task offloading possible to save energy for mobile devices. \nWe achieved a great deal of progress toward developing a realistic offloading framework. \nFirst, we examined the feasibility of exploiting the offloading technique to save mobile device energy using the cloud as the place to execute the task instead of executing it on the mobile device. \nOur evaluation study reveals that the offloading does not always save energy; in cases where the energy for the computation is less than the energy for communication no energy is saved. \nTherefore, the need for the offloading decision is vital to make the offloading beneficial. \nSecond, we developed mathematical models for the energy consumption of a mobile device and its applications. \nThese models were then used to develop mathematical models that estimate the energy consumption on the networking and the computing activities at the application level. \nWe modelled the energy consumption of the networking activity for the Transmission Control Protocol (TCP) over Wireless Local Area Network (WLAN), the Third Generation (3G), and the Fourth Generation (4G) of mobile telecommunication networks. \nFurthermore, we modelled the energy consumption of the computing activity for the mobile multi-core Central Processing Unit (CPU) and storage unit. \nThird, we identified and classified the system parameters affecting the offloading decision and built our offloading framework based on them. \nIn addition, we implemented and validated the proposed framework experimentally using a real mobile device, cloud, and application. \n \nThe experimental results reveal that task offloading is beneficial for mobile devices given that in some cases it saves more than 70% of the energy required to execute a task. \nAdditionally, our energy models accurately estimate the energy consumption for the networking and computing activities. \nThis accuracy allows the offloading framework to make the correct decision as to whether or not offloading a given task saves energy. \nOur framework is built to be applicable to modern mobile devices and expandable by considering all system parameters that have impact on the offloading decision. \nIn fact, the experimental validation proves that our framework is practical to real life scenarios. \nThis framework gives researchers in the field useful tools to design energy efficient offloading systems for the coming years when the offloading will be common.
Source: UWSpace (University of Waterloo)
Author: Majid Altamimi
Institution: University of Waterloo
Language: English
Topic: Green IT and Sustainability
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2264518842
