Access networks are continuously going through many reformations to make them better suited for various demanding applications and meet new challenging requirements. On one hand, incorporating fog and edge computing has become a necessity for alleviating network congestions and supporting numerous applications that can no longer rely on the resources of a remote cloud. On the other hand, energy-efficiency has grown to be essential for these networks to reduce both their operational costs and carbon footprint but often leads to degradation in their network performance. In this paper, we study the challenges posed by these two imperatives by examining the integration of fog computing with passive optical networks (PONs) under power-conserving frameworks. As most power-conserving frameworks in the literature are centralized-based, we also propose a decentralized-based framework and compare its performance with its centralized counterpart. We study the possible cloudlet placements and the offloading performance in each allocation paradigm to determine which paradigm is able to meet the requirements of next-generation access networks by having better network performance with less energy consumption.

Power conservation in passive optical networks (PONs) has been an active area of research since the cyclic sleep-mode was first proposed for optical network units (ONUs). Many studies have then settled upon locking downstream and upstream transmissions for each ONU in a cyclic fixed slot, thus allowing the ONU to switch to sleep-mode for the rest of the transmission cycle. However, such fixed allocation limits the flexibility and dynamicity of the bandwidth allocation and leads to upstream underutilization. Moreover, to maximize power conservation, the cycle duration used must be long enough to make up for mode-switching overheads, which significantly degrades the network performance in terms of packet delays. In this paper, we develop a novel energy-efficient framework for Ethernet PONs (EPONs). To that end, we propose different upstream allocation schemes to improve the fixed-slot performance while maintaining energy-efficiency at acceptable levels. We also propose a more accurate arrangement for downstream-upstream locking. Moreover, we use a long-reach PON setting, where the long propagation delays impact the network performance posing further challenges to the bandwidth allocation. Numerical results show that, under heavily loaded network conditions, packet delays can be reduced by around 60% at an additional power cost of less than 5%.

Integrating optical access networks with fog computing combines the high capacity of optical fiber with closer-to-the-edge computing and storage capabilities. Such integration is believed to form a highly capable backhaul that will alleviate network congestions, serve local demands with less energy consumption, and live up to the requirements of tomorrow's access networks and application requirements. This integration however requires reexamining the bandwidth allocation in addition to reconsidering the network architecture itself, which was not designed to support direct edge-to-edge communications. Moreover, the growing demands for energy-efficient access networks also add the requirement for sleep-aware bandwidth allocation and a power-conserving framework. In this paper, we study the offloading performance in a long-reach passive optical network (LR-PON) when the underlying bandwidth allocation is either centralized or decentralized. Moreover, we investigate how offloading can be supported in each paradigm when optical network units (ONUs) go through a cyclic sleep-mode to conserve energy. We consider this paper to be one of the first to look into integrating fog with LR-PONs under a power-conserving framework and examine which allocation paradigm would be fit to carry offloaded traffic with better network performance and energy-efficiency within this new setting.

Several centralized algorithms have been proposed for bandwidth-allocation in passive optical networks (PONs); making the optical line terminal (OLT) in the central office the intelligent device that arbitrates time-division access to the shared upstream channel. When the distance between the OLT and optical network units (ONUs) is extended from 20 km to 100 km and beyond, as suggested by next generation long-reach PONs, it becomes difficult for centralized algorithms to maintain their performance and to support service differentiation for real-time applications. This is because these algorithms are based on bandwidth negotiation messages frequently exchanged between the OLT and ONUs, which become seriously delayed, thereby delaying packet transmissions and causing the performance to degrade. Decentralized schemes, on the other hand, are independent of the OLT distance and can perform better with such extended reach. In this paper, we introduce and enhance the performance of a decentralized scheme for the emerging LR-PON. Simulation results show that the proposed scheme can significantly reduce upstream packet delays below those of centralized algorithms by 80% while keeping a high throughput.