Date of Award

December 2019

Document Type


Degree Name

Doctor of Philosophy (PhD)


School of Computing

Committee Member

Jim Martin

Committee Member

Brian Dean

Committee Member

Kuang-Ching Wang

Committee Member

James Westall


Mobile wireless networks are always challenged by growing application demand. The increasing heterogeneity of both mobile device connection capability and wireless network coverage forms a general heterogeneous wireless network (HetNet). This type of HetNet contains sub-networks of different Radio Access Technologies. How to better coordinate the mappings of flows between Access Points (AP) and User Equipment (UE) inside this type of HetNet to improve system and user-level performance is an interesting research problem. The flow mapping systems used by off-the-shelf mobile devices make policy-based decisions from local information. Several global information based flow mapping systems that use Generalized Proportional Fairness (GPF) as the optimization objectives have been proposed to improve the system-level performance. However, they have not been compared with both the local-policy based approaches and the optimal solution under the same assumptions with variations of system parameters. Therefore, it is still unclear to the community whether it is worthwhile to construct a flow mapping system for HetNets composed by LTE and WiFi networks, even under a simplified assumption of only optimizing throughput related system performance metrics. In this dissertation, we evaluate three types of flow mapping systems: Global Information based Flow Mapping Systems (GIFMS), Local Information based Flow Mapping Systems (LIFMS), and Semi-GIFMS. We evaluate these systems with metrics related to both the spectrum efficiency and flow-level fairness under the following variations of system parameters: 1) topologies of UEs; 2) coverage of APs; 3) number of UEs; 4) number of non-participating UEs; 5) on-off session dynamics; 6) UE mobility. We also discuss options to implement each type of flow mapping systems and any relevant trade-offs.

From the evaluations, we find that the currently-in-use WiFi preferred local greedy flow mapping system provides far poorer spectral efficiency and generalized proportional fairness than all the other tested flow mapping systems, including the local greedy flow mapping systems that give LTE and WiFi equal opportunities (local-greedy-equal-chance) in most settings. This finding indicates that the flow mapping system in use has much room for improvement in terms of GPF and aggregate throughput. The performance of local-greedy-equal-chance is close to that of the global and AP-level information based systems under some UE topologies. However, their performance is not as consistent as the global and AP-level based systems when UEs form clusters that produce AP load imbalance.

We also derive the incremental evaluations of GPF for both proportional and max-min fair scheduled APs. Based on these derivations, we propose a design for AP-level information based flow mapping system or Semi-GIFMS. It is an event-triggered flow mapping system based on minimum AP-level metrics monitoring and dissemination. From our evaluation and analysis, this flow mapping system performs equivalent to or better than GIFMS in terms of both GPF and aggregate throughput in all the tested scenarios. It also owns the advantages of lower overhead and not requiring an additional scheduling server. We think it is the best choice for the next generation HetNets where APs can be modified to monitor and broadcast the minimum information identified.

Furthermore, we find that the number of UEs, number of non-participating UEs, coverage of APs, bandwidth sharing types of APs, on-off session and UE mobility dynamics do not have a major impact on the relative performance difference among various flow mapping systems.



To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.