In recent years, the number of devices accessing wireless services has increased exponentially; as a result, a large number of often heterogeneous devices compete for access to the same limited available spectral resources in order to satisfy their traffic requirements. This competition could not have been more evident than in the unlicensed 2.4/5 GHz bands. Despite new spectrum bands (e.g., 60 GHz band) be- coming available, the benefits (e.g., low-cost access, extended range, non-LoS communication) of operating in the 2.4/5 GHz unlicensed bands are expected to keep their popularity intact. Consequently, the critical challenge to all the network service providers that cater wireless services in these bands is to find novel ways to judiciously manage the available (unlicensed) spectral resources, while ensuring that: (i) all users get sufficient bandwidth to satisfy their traffic demand and (ii) there is no wastage/under-utilization of the available spectral resources.

In this dissertation, we identify two popular types of wireless net- work deployments that often suffer significant performance degradation, and can benefit from efficient spectral resource provisioning in the unlicensed band. We then propose novel solutions to optimize their spectral resource utilization and improve their overall network performance. Our proposed solutions are backed by strong theoretical frameworks, yet practically-implementable, standard-compliant, and readily deployable over the existing wireless network stack without any changes.

In the first part of the dissertation, we focus on WLAN deployments in large enterprises and high-density venues, where thousands of devices, many running high bandwidth applications, simultaneously access the available Wi-Fi services. We show that the traditional WLAN architecture that provides static spectral resource assignment is ill-equipped to handle occasional surges in the user-traffic, and more importantly suffers due to the spatial-temporal changes in the traffic patterns. To alleviate these problems, we propose AmorFi, an amorphous WLAN architecture with the ability to dynamically (re-)provision the available spectrum resources to address traffic surges and the spatial-temporal changes in the user-traffic demands. Our evaluations show that AmorFi can guarantee 1.5x-2x improved throughput performance than the traditional WLANs.

In the second part of the dissertation, we focus on the recently introduced LTE deployments in the unlicensed spectrum. Since LTE was envisioned to operate in the licensed band, free from any external interference, it was never designed to share its available spectral resources with other radio access technologies. However, its recent advent into the unlicensed band, where it shares the spectrum with the incumbent Wi-Fi, creates a conflict between the asynchronous interference present in the unlicensed band and LTEs scheduled, and synchronous channel access, that can result in massive under-utilization/wastage of the spectral resources. Consequently, we propose BLU and ELI, two novel systems that aim to empower LTE networks operating in the unlicensed spectrum with interference awareness to ensure optimal utilization of the available spectral resources. BLU and ELI individually provide up to 2x improved performance when compared to existing interference-agnostic LTE networks.