Emerging modular cable network architectures distribute some cable headend functions to remote nodes that are located close to the broadcast cable links reaching the cable modems (CMs) in the subscriber homes and businesses. In the Remote- PHY (R-PHY) architecture, a Remote PHY Device (RPD) conducts the physical layer processing for the analog cable transmissions, while the headend runs the DOCSIS medium access control (MAC) for the upstream transmissions of the distributed CMs over the shared cable link. In contrast, in the Remote MACPHY (R-MACPHY) ar- chitecture, a Remote MACPHY Device (RMD) conducts both the physical and MAC layer processing. The dissertation objective is to conduct a comprehensive perfor- mance comparison of the R-PHY and R-MACPHY architectures. Also, development of analytical delay models for the polling-based MAC with Gated bandwidth alloca- tion of Poisson traffic in the R-PHY and R-MACPHY architectures and conducting extensive simulations to assess the accuracy of the analytical model and to evaluate the delay-throughput performance of the R-PHY and R-MACPHY architectures for a wide range of deployment and operating scenarios. Performance evaluations ex- tend to the use of Ethernet Passive Optical Network (EPON) as transport network between remote nodes and headend. The results show that for long CIN distances above 100 miles, the R-MACPHY architecture achieves significantly shorter mean up- stream packet delays than the R-PHY architecture, especially for bursty traffic. The extensive comparative R-PHY and R-MACPHY comparative evaluation can serve as a basis for the planning of modular broadcast cable based access networks.