Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.678181
Title: User-experience-aware system optimisation for mobile systems
Author: Bischoff, Alexander S.
ISNI:       0000 0004 5370 1863
Awarding Body: University of Southampton
Current Institution: University of Southampton
Date of Award: 2016
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
Abstract:
This thesis considers the concept of Quality of Experience (QoE) in the context of mobile electronic consumer devices, such as smartphones. The modern smartphone is expected to deliver a high level of user experience across a wide variety of tasks, whilst remaining as power efficient as possible. Commonly, mobile devices undergo runtime optimisation to achieve the required level of performance, with the energy consumption being a secondary concern. In this thesis, we stress that it is vital to not focus on the raw performance of the device, but instead to concentrate on the needs and desires of the end user. This approach ensures that the end-user is satisfied at all times, and that the power consumption for a given level of user experience is minimised. Hence, we advocate user-experience-aware system optimisation. We introduce the concept of Quality of Experience, which has traditionally been used only in the telecommunications industry, to mobile system optimisation. We develop user experience models in the form of utility functions, and use these to translate lowlevel metrics into the delivered user experience. Upon these models we build simple, yet effective, QoE-aware Central Processing Unit (CPU) and Graphics Processing Unit (GPU) governing algorithms which adjust the performance and power consumption at runtime to meet user experience requirements. When creating our algorithms, we first analyse and characterise the operation of both CPU and GPU workloads. Specifically, we investigate how the level of compute-boundedness or memory-boundedness of CPU workloads affects frequency scalability, as well as determining how the available bandwidth and core count for a GPU affects the rendering performance. We combine both gem5-based simulation driven analysis and hardware-based verification in order to validate our QoE-aware governing algorithms. Additionally, we validate the operation of our algorithms using a variety of common mobile workloads. As part of this work, we have also extended the gem5 simulator to allow use to investigate the potential for finegrained Dynamic Voltage and Frequency Scaling (DVFS) adjustment, and use this as a platform to investigate the operation of the Linux CPUFreq governors used on modern mobile platforms.
Supervisor: Al-Hashimi, Bashir Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.678181  DOI: Not available
Share: