Programmable Scheduler Architectures for Multi-Tenant Wireless Networks

In future wireless, WiFi or cellular networks, physical networks are shared among multiple virtual networks. Each virtual network becomes a tenant in the physical network, which is known as a multi-tenant network. Operators need to handle time-varying network conditions and various traffic demands of different tenants by programming the radio resource scheduler logic of edge base stations (BSs). In this thesis, we first use an existing centralized programmable scheduler architecture for WiFi networks to develop a radio resource scheduler that allocates spectrum and service set identifiers (SSIDs) used by BSs in WiFi, in order to improve the throughput of multi-tenant WiFi networks. Our experiments indicate that the network throughput with the proposed scheduler remains the same with an increasing number of tenants, in contrast with other existing schedulers that show a linear decrease in the network throughput. The above centralized programmable scheduler architecture is not able to program the logic that schedules the transmissions of every radio frame, which is critical to satisfying the flexible quality of service (QoS) demands of different tenants. To address this issue, we then propose a new distributed real-time programmable scheduler architecture, where we refer to a real-time scheduler as one with logic running at every transmission time interval (TTI). In the proposed architecture, the scheduling logic is written in the controller, independently of the underlying cellular BS software, and is then executed in real-time at the edge BS with the help of a scheduler agent. The hardware prototyping activities with standards-compliant setup validate the practicality of the proposed architecture. Experimental results show that the proposed architecture allows the operator to apply real-time customized strategies according to the needs of the network tenants