Enabling mobile network densification
- Demand for mobile traffic is proliferating. With increments in spectrum and spectral efficiency not able to keep pace with this proliferation, mobile networks inevitably need to densify in order to meet this increasing demand. However, densification of mobile networks has been impeded in practice by two broad challenges - interference and congestion. These challenges need to be solved in order to enable mobile network densification at scale. Interference is a link-layer challenge. Managing interference in dense mobile networks requires sub-millisecond coordination between every pair of neighboring cells whose coverage areas overlap. This is needed so they can dynamically adapt their scheduling and transmission parameters such that their transmissions do not interfere. While macro cells are equipped with backhaul solutions for such low-latency coordination, existing approaches for backhauling small cells either do not provide such low latencies or require deployment of expensive infrastructure, neither of which is a practical solution for solving this problem at scale. I present QuickC, a practical sub-millisecond coordination link for small cells. QuickC is built on the insight that the small cells are deployed in the coverage areas of macro cells, and a low-latency wireless link for coordination could potentially be built between small cells and macro cells over the licensed spectrum. The challenge however is that licensed spectrum is expensive. I show that it is possible to design a low-latency coordination link over the licensed spectrum that negligibly affects the performance of the existing users of the spectrum, which means that this link can be built at (almost) no spectrum cost. I also show that this link can enable dynamic coordination between neighboring cells to tackle congestion and consequently improve the capacity of dense mobile networks. Congestion is a network-layer challenge. Managing congestion, or more generally managing connectivity and traffic delivery, in dense mobile networks requires the ability to answer what-if questions about how the future performance of clients at the granularity of individual sessions will be impacted if a certain network-control action is taken, like admitting a new client or handing over a client from one cell to another. This is needed to proactively decide what network-control action is best suited in any scenario to avoid congestion and deliver a good quality of experience to clients. Existing approaches either do not have this ability or have this ability only for much-coarser granularities or require invasive changes to existing client devices to obtain fine-grained performance logs, neither of which is a practical solution that can solve this problem at scale. I design ForeC, a predictive analytics system for mobile networks. ForeC is built on the insight that the limited statistics exposed by the cells for monitoring their health and performance contain information at sufficient granularity to view the performance of clients and cells. The challenge however is to discover how cellular networks can be forecasted and how the performance of clients can be predicted. I show how ForeC solves these challenges and enables a wide-variety of data-driven network optimizations that tackle congestion to improve the quality of experience of clients and the overall utilization of network resources.
|Type of resource
|electronic; electronic resource; remote
|1 online resource.
|Stanford University, Department of Electrical Engineering.
|Statement of responsibility
|Submitted to the Department of Electrical Engineering.
|Thesis (Ph.D.)--Stanford University, 2016.
- © 2016 by Rakesh Misra
- This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).
Also listed in
Loading usage metrics...