• Sadia Saleem

Student thesis: Phd


Ad-hoc Mobile Collaborative Community (MCC) enables two or more low band- width mobile phone/PDA network channels to achieve a virtual high-bandwidth channel for collaborative data transfer using inverse multiplexing techniques. This research looks at how to achieve best effort QoS for MCC communities and pro- poses a novel Risk-Aware Workload (RAW) scheduler, to support efficient collab- orative data transfer in MCC. To this end, the thesis has made the following four novel contributions. Firstly, it presents a comprehensive requirements analysis of the MCC scenarios and literature review of the existing solutions proposed to address these requirements. The analysis and literature review have led to the identification of our research aim, i.e. to design the RAW scheduler, an energy and QoS performance efficient workload scheduler for MCC. Secondly, it presents a novel architecture for the RAW scheduler. The architecture is designed for battery and memory constrained mobile devices and supports best effort QoS for multi-stream applications. The design has made use of a modular approach so that any of the architectural modules can be replaced as technologies advance without affecting other modules. Thirdly, it presents the design and evaluation of an energy-efficient Multi-level Work Queue (MWQ) scheduling algorithm, a core component of the RAW scheduler. The algorithm is application require- ment aware and is adaptive to the changes in the underlying channel conditions. We have evaluated the QoS performance of the MWQ scheduling algorithm by comparing it against the performances of three other state-of-the-art scheduling algorithms using the NCTUNs simulator. The simulation setup aggregates the bandwidth of two independent channels and results show an improvement of 45% in total data transfer time for MWQ as compared to the Single Work Queue (SWQ) algorithm and up to 62% as compared to the Round Robin (RR) algo- rithm. In comparison with the Channel Pinned (CP) algorithm, the improvement ranges between 28% to 95%, depending on the distribution of data to the two14channels. Fourthly, we have implemented an "HTTP Downloader" application with the novel RAW scheduling design built-in. This test-bed implementation is done using real world mobile devices and network communication technologies. We performed a detailed evaluation of the QoS performance of our "HTTP Down- loader". The evaluation results show that the RAW scheduler delivers better QoS results in terms of average throughput and total data transfer times by adapting to the changes in the channel conditions. It has been found that, among the four alternatives, the MWQ algorithm delivers the best results when we plugged it into our test-bed application. To download a 2MB file from the Internet using four collaborators, MWQ showed an improvement of 46% in total download time as compared to the RR algorithm. The percentage improvement values in the similar scenario for the CP algorithm is 35% and the SWQ algorithm is 17%.
Date of Award1 Aug 2017
Original languageEnglish
Awarding Institution
  • The University of Manchester
SupervisorNing Zhang (Supervisor)


  • workload scheduling
  • collaborative data transfer
  • mobile communities

Cite this