Recent Advances in Fog Radio Access Networks: Performance Analysis and Radio Resource Allocation

I. INTRODUCTION

• C-RAN centralizes baseband processing in the BBU pool, while constrained fronthaul capacity and delay degrade SE and EE and centralized processing burdens the pool.Source: §I · p. 1
• F-RANs move some processing, resource management, and storage to RRHs and UEs, alleviating fronthaul and BBU burdens and reducing latency.The article describes F-RAN as a socially aware 5G architecture intended to address disadvantages of C-RAN.Source: §I · p. 1
• Prior work covers F-RAN architecture and caching, but outage-probability and ergodic-capacity analysis remain difficult, constraining development.The article identifies performance analysis and radio-resource allocation as key issues and surveys recent advances in both.Source: §I · p. 1

II. CACHE-BASED PERFORMANCE ANALYSIS

• The cloud center includes a content cache, BBU pool, and controller; RRHs connect to the BBU pool via fronthaul, and the HPN connects to the content cloud via backhaul.Source: §II · p. 2
• EC-APs combine front radio frequency with local collaborative signal processing, cooperative radio-resource management, and caching.Collaboration among adjacent EC-APs and D2D devices can reduce fronthaul overload and alleviate queuing and transmission latency.Source: §II · p. 2
• When CRSP and CRRM functions move to the BBU pool, an EC-AP becomes a traditional RRH; EC-UEs are UEs accessing EC-APs and working in D2D mode.Source: §II · p. 2

A. Edge caching impacting on the performance of F-RAN

• Collaborative infrastructure-and-device caching assigns high-rank popular social media to EC-APs and smaller popular items to D2D sharing, reducing latency.The distribution uses content popularity and size to increase successful delivery probability.Source: A. Edge caching impacting on the… · p. 2, Figure 2 · p. 2
• Popularity-aware caching performs best among the compared caching strategies because EC-APs are more likely to cooperate and achieve cooperation gain.Random caching only outperforms no caching, while probabilistic caching falls between the two; the comparison is shown in Fig. 3(a).Source: A. Edge caching impacting on the… · p. 2, Figure 3 · p. 3
• EE increases as local cluster cache size grows, and the cited comparison reports F-RAN outperforming C-RAN due to edge caching.Fig. 3(b) compares the nested coalition formation game-based algorithm with its suboptimal alternative.Source: A. Edge caching impacting on the… · p. 2, Figure 3 · p. 3
• A nested coalition formation game selects cell associations using jointly considered traditional radio-resource and caching power, with a suboptimal algorithm designed to reduce computational complexity.The cited work defines energy consumption as the product of the power of keeping a content object and its size.Source: A. Edge caching impacting on the… · p. 2

B. Mode Selection impacting on the performance of F-RAN

• F-RAN UEs can choose D2D, EC-AP, local distributed coordination, or global C-RAN access according to nearby content availability.D2D requires a sufficiently close neighboring UE with the desired content; coordination applies when multiple EC-APs can provide it.Source: B. Mode Selection impacting on the… · p. 3
• Local D2D or EC-AP delivery is preferred over backhaul or fronthaul access for high SE/EE and low latency.Source: B. Mode Selection impacting on the… · p. 3
• Adaptive model selection in F-RANs can improve SE and decrease latency.Source: B. Mode Selection impacting on the… · p. 3
• Mode selection should account for cached content, desired SINR, and energy consumption; sufficiently short D2D distance reduces energy use.The cited comparison attributes the reduction to higher SINR, lower transmit power, and lower D2D transmitter circuit power.Source: B. Mode Selection impacting on the… · p. 3
• Local VB (D2D) has the best latency, subVB (distributed cooperation) and VB (nearest access point) are similar with subVB better, and outside (global C-RAN) is slowest.Outside mode incurs Internet retrieval and fronthaul transmission time.Source: B. Mode Selection impacting on the… · p. 3

Radio Resource Allocation

• F-RAN resource allocation jointly designs precoding, resource blocks, scheduling, and cell association to optimize spectral efficiency, energy efficiency, and latency.The deployment of edge-caching EC-APs and D2D UEs makes allocation more complex than in C-RANs.Source: §III · p. 3

Spectral Efficiency Optimization

• F-RAN cell association favors EC-APs caching the requested content, subject to sufficient received signal strength; scheduling and power allocation then optimize spectral efficiency.F-RAN spectral efficiency must be balanced against transmit power and fronthaul capacity.Source: §III · p. 4
• Larger EC-AP cache sizes and clusters with more serving EC-APs are associated with higher ergodic rates for local distributed coordination.The analysis considers node locations, cache sizes, and user access modes.Source: §III · p. 4

Energy Efficiency Optimization

• Larger cluster cache sizes nearly triple F-RAN average effective energy efficiency by letting more UEs obtain cached content locally.Local delivery avoids fronthaul and backhaul, unlike centralized C-RAN content delivery.Source: §III · p. 4

Latency Optimization

• A cache-capacity-constrained allocation algorithm nearly matches exhaustive-search download-delay performance, while larger cache sizes sharply reduce download delay.Local EC-AP delivery can reduce fronthaul delay, but EC-AP-to-UE latency remains relevant.Source: §III · p. 5, Figure 6 · p. 5

• Open challenges remain in F-RAN performance analysis and radio resource allocation, particularly for 5G architecture and social awareness.Source: §IV · p. 5

A. F-RAN architecture for 5G

• F-RANs are described as compatible with other 5G systems and able to use massive MIMO, cognitive radio, millimeter-wave communications, and NOMA directly.Source: A. F-RAN architecture for 5G · p. 5
• F-RANs are anticipated to support IoT applications including vehicular connectivity, smart cities, and industrial automation, while meeting varied SE, EE, latency, and reliability needs.The architecture is described as offering local caching, real-time CRSP, flexible edge CRRM, rapid scaling, and reduced fronthaul and BBU burdens.Source: A. F-RAN architecture for 5G · p. 5

B. Social-awareness F-RAN

• Social relationships affect D2D communication success and contribute to F-RAN spectral efficiency, energy efficiency, and latency, but are seldom considered in analysis or allocation.Users tend to exchange content within communities and avoid communicating with unfamiliar users; the paper calls for social-aware F-RAN research.Source: B. Social-awareness F-RAN · p. 5

V. CONCLUSION

• F-RAN performance analysis and radio resource allocation are more complex but advanced than in C-RANs because of local edge caching and adaptive model selection.The survey identifies SE, EE, and latency as the three major issues it examines.Source: §V · p. 5
• The survey summarizes numerical results on how caching and model-selection strategies affect SE, EE, and latency.Source: §V · p. 5
• Because F-RAN-based 5G remains an infant field, the survey calls for further investigation, particularly of F-RAN architecture and social-awareness F-RAN.It anticipates transferring the discussed performance-analysis and resource-allocation issues to standards organizations.Source: §V · p. 5