PaperlayerView exact revision on arXiv

Source-linked AI summary

Fog Computing based Radio Access Networks: Issues and Challenges

Mugen Peng, Shi Yan, Kecheng Zhang, Chonggang Wang

arXiv:1506.04233v1cs.ITcs.NI

TL;DR

F-RANs address constrained fronthaul and centralized BBU processing in 5G by distributing radio processing, resource management, and storage across edge devices. The paper presents the architecture and key techniques, including transmission-mode selection and interference suppression, and identifies remaining open issues. Its supported conclusion is that F-RANs incorporate fog computing into H-CRANs while shifting cooperative functions toward the edge.

  • Problem

    F-RANs target challenges associated with constrained fronthaul and centralized BBU processing in 5G radio access networks.

  • Method

    The paper presents an F-RAN architecture with F-APs and F-UEs, together with cooperative radio processing, resource management, caching, transmission-mode selection, and interference suppression.

  • Results

    The paper reports that shifting cooperative radio signal processing and resource management to edge devices alleviates fronthaul and BBU-pool burden, while local caching enables packet-service offload.

  • Takeaways & Limitations

    F-RANs provide a distributed architecture that combines fog computing with H-CRANs and supports local cooperative processing and storage.

Abstract

from arXiv · show

A fog computing based radio access network (F-RAN) is presented in this article as a promising paradigm for the fifth generation (5G) wireless communication system to provide high spectral and energy efficiency. The core idea is to take full advantages of local radio signal processing, cooperative radio resource management, and distributed storing capabilities in edge devices, which can decrease the heavy burden on fronthaul and avoid large-scale radio signal processing in the centralized baseband unit pool. This article comprehensively presents the system architecture and key techniques of F-RANs. In particular, key techniques and their corresponding solutions, including transmission mode selection and interference suppression, are discussed. Open issues in terms of edge caching, software-defined networking, and network function virtualization, are also identified.

I. INTRODUCTION

The paper proposes F-RANs to address fronthaul and centralized-processing constraints in 5G by moving radio processing, resource management, and storage toward edge devices. It presents the architecture, adaptive transmission and interference techniques, and open research issues.

  • 5G targets at least 1000-fold system capacity growth and at least 10-fold energy-efficiency growth over 4G.
  • C-RAN and H-CRAN architectures face fronthaul capacity constraints, redundant traffic, centralized-processing burdens, and inefficient provisioning for peak demand.
  • F-RAN extends fog computing into RANs by exploiting local caching, collaborative radio signal processing, cooperative radio resource management, and distributed storage at RRHs and smart UEs.
  • The paper defines F-APs as fog-enabled access points and discusses adaptive transmission-mode selection and interference suppression for F-RANs.
  • It identifies edge caching, software-defined networking, and network function virtualization as future challenges and open issues.

II. F-RAN SYSTEM ARCHITECTURE

The F-RAN architecture distributes communication, processing, control, and storage functions across fog-layer devices while retaining centralized cloud resources. Local caching and processing reduce traffic and processing burden on the fronthaul and BBU pool.

  • Architecture evolution: Local delivery from adjacent F-APs or F-UEs can save constrained fronthaul spectral usage and decrease transmission delay.
  • Architecture evolution: F-RANs combine centralized and distributed clouds, including centralized communication and storage, centralized control, distributed communication, and distributed storage clouds.
  • Architecture evolution: F-APs and F-UEs host local radio signal processing and cooperative radio resource management, while centralized clouds retain broader system functions.
  • System model: F-UEs communicate through device-to-device or F-UE relay modes, while F-APs and HPNs provide network access and control-plane connectivity.
  • Architecture benefits: Shifting CRSP and CRRM to F-APs and F-UEs alleviates fronthaul and BBU-pool burden, while edge caching enables packet-service offload from centralized caching.

A. Key Components: F-APs and F-UEs

F-APs and F-UEs extend conventional access points and user equipment with fog capabilities for local processing, cooperation, caching, and flexible transmission. These functions reduce dependence on centralized processing and fronthaul links.

  • Key components: F-APs integrate radio-frequency functions with local distributed CRSP and simple CRRM, while F-UEs are user devices operating in the fog layer.
  • F-AP functions: F-APs process local CRSP and CRRM, suppress interference, support D2D spectral sharing, and compress-and-forward information through fronthaul.
  • F-AP cooperation: Collaborative processing among adjacent F-APs releases fronthaul overload and alleviates queuing and transmission latency.
  • Cooperative processing: Distributed CRSP among adjacent F-APs can provide diversity and multiplexing gains for F-UEs without consuming fronthaul links.
  • Interference and adaptation: Interference suppression can use coordinated multipoint transmission and reception, while centralized CRSP and CRRM remain available when local processing is insufficient.

B. Hierarchical Architectures

F-RANs distribute radio processing, coordination, and storage across fog, cloud, access, and terminal layers. This architecture reduces fronthaul constraints while supporting centralized and local processing options.

  • Architecture: The hierarchical F-RAN architecture combines fog computing and cloud computing layers with network-access and terminal layers.The cloud layer provides centralized computing and caching, while fog and terminal entities support distributed processing and communication.
  • Terminal layer: D2D communication reuses radio resources with F-UEs connecting to F-APs, supporting high-data-rate transmission and potentially enhancing overall throughput.Its use is constrained by communication distance, F-UE capabilities, and the lack of D2D support in traditional UEs.
  • Network access layer: HPNs deliver control signalling, reference signals, seamless coverage, and basic bit rates for highly mobile F-UEs.This arrangement can decrease unnecessary handovers and alleviate synchronous constraints.
  • Network access layer: Adjacent F-APs use interconnected mesh or tree-like topologies for distributed CRSP, CRRM, interference suppression, and local packet delivery.The connections use data and control interfaces, and distributed processing operates without assistance from the BBU pool.
  • Network access layer: The tree-like topology has about 50 percent lower wireless cluster feasibility than the mesh topology, with significantly reduced network deployment and maintenance cost.Both topologies can decrease negative influences from capacity-constrained fronthaul links; the tree-like topology is preferred in practical F-RANs.
  • Distributed processing and storage: F-APs and F-UEs execute many CRSP and CRRM functions locally, while edge devices store packet traffic instead of sending all processing and storage to the cloud.When F-APs are simplified into traditional RRHs, received signals are forwarded to the centralized BBU pool for large-scale processing.

III. TRANSMISSION MODE SELECTION

F-RANs adaptively select among four transmission modes according to mobility, user distance, QoS, device capabilities, content location, and available local performance. The policy ranges from direct D2D or relay communication to local coordination, global C-RAN processing, or HPN access.

  • Mode set: The four transmission modes are D2D and relay, local distributed coordination, global C-RAN, and HPN mode.These modes allow F-UEs to access the F-RAN adaptively according to operating conditions.
  • Selection procedure: The desired F-UE selects its mode under HPN supervision using estimated movement speed and paired-user distance from HPN broadcast pilot channels.Selection also accounts for QoS requirements, location, and processing and caching capabilities.
  • Selection rules: HPN mode receives priority for high-speed mobility or real-time voice communication and provides seamless coverage with basic QoS support.It can also decrease control-channel overhead and avoid unnecessary handovers.
  • Selection rules: D2D mode is triggered for slowly moving paired F-UEs whose distance is no greater than threshold D1.For distances between D1 and D2, an adjacent F-UE can provide relay communication when it offers better performance than other modes.
  • Selection rules: Local distributed coordination is selected for slowly moving F-UEs at intermediate distances or when at least one paired F-UE lacks D2D and relay support.The global C-RAN mode is used when local coordination cannot meet expected performance, distance exceeds D3, or content comes from the cloud server.

A. D2D and Relay Mode

The D2D and relay mode provides user-centric terminal-layer communication through direct D2D or wireless relay links. Its rate gains depend on channel conditions and decline when D2D-user density becomes sufficiently large.

  • Mode operation: The D2D and relay mode lets paired F-UEs communicate directly or through a wireless relay, without relying on centralized radio processing.The HPN assigns device identification for each F-UE in this mode.
  • Benefits: D2D and relay techniques can achieve significant performance gains and effectively relieve fronthaul burden because processing occurs in the terminal layer.D2D reuses radio resources and is beneficial for high-data-rate transmission.
  • Limitations: Spatial average rate performance is severely degraded in every scenario when D2D-user density λD becomes sufficiently large.The mode is also constrained by communication distance and F-UE capability, and traditional UEs without D2D support cannot use it.

B. Local Distributed Coordination Mode

Local distributed coordination adaptively forms F-AP clusters to coordinate interference when local content and interconnections make distributed processing feasible. When these conditions fail, global C-RAN or HPN operation is selected.

  • Local coordination: Local distributed coordination is used when traffic comes from F-APs rather than the cloud server, with interference-coordination gains mainly arising from CoMP.The corresponding F-AP cluster is formed adaptively by considering implementation complexity and CoMP gains.
  • Local coordination: CoMP gains depend strictly on the F-RAN cluster topology and the backhaul capacity connecting F-APs.Thus, the coordination design must account for both network connectivity and link capacity.
  • Performance: Cell-edge spectral efficiency increases by about 70% in the downlink and 122% in the uplink under local distributed coordination.F-AP association is also identified as critical to improving spectral efficiency.
  • Global C-RAN fallback: If coordinating F-APs are not interconnected or content exists only on the cloud server, the global C-RAN mode is adopted.RRHs forward received signals to the BBU pool, which performs CRSP and CRRM centrally; several RRHs can jointly serve the desired UE.
  • Mode selection consequences: The other three modes significantly decrease fronthaul capacity demands, alleviating fronthaul capacity and latency constraints.HPN mode is preferred for bursty low-volume content or mobility beyond a predefined threshold, while providing basic QoS and seamless coverage.
  • HPN interference management: Enhanced soft fractional frequency reuse can mitigate inter-tier interference between HPNs and F-APs, with reported spectral-efficiency and energy-efficiency gains.The cited S-FFR scheme allocates only partial resources to low-QoS UEs and reserves remaining resources for high-QoS F-AP users.

IV. INTERFERENCE SUPPRESSION

F-RAN interference suppression combines coordinated precoding and coordinated scheduling across centralized and distributed coordination modes. The section also evaluates clustering and utility-function settings through energy-efficiency and data-rate trends.

  • Interference suppression is categorized into coordinated precoding and coordinated scheduling for F-RANs.
  • Coordinated precoding decreases interference centrally for global C-RAN operation and distributively for local coordination.
  • Distributed joint-processing CoMP uses sparse precoding to reduce processing complexity and channel-estimation overhead.
  • Algorithm 1 forms intro-cluster cooperation groups through merge-and-split coalitional formation, evaluated against grand-cluster and no-clustering baselines.
  • With fixed power consumption, energy-efficiency curves nearly match corresponding average-data-rate curves; τ = 0.1 mitigates the power-consumption component and allows flexible cluster sizes.
  • As SINR threshold increases, average data rate rises in lower and medium regions but grows more slowly or declines at high thresholds because successful access probability decreases.

B. Coordinated Scheduling

Coordinated scheduling addresses interference among F-UEs using centralized access control and delay-aware optimization tools. The reported comparison shows that cellular success probability varies with spectrum occupation and that COAC outperforms DRAC.

  • Coordinated scheduling mitigates interference in the MAC layer, including interference among F-UEs using different transmission modes.
  • A stochastic gradient algorithm dynamically allocates power and transmission rate with low computing complexity and robustness to random traffic and imperfect CSIT.
  • COAC lets D2D F-UEs access sub-channels opportunistically under centralized HPN control, whereas DRAC is distributed random access control.
  • Cellular success probability is evaluated across sparse, medium, and dense D2D densities because DRAC and COAC have the same D2D success probability.
  • Cellular success probabilities decrease as ε increases, with ε = 0 and ε = 1 matching the upper and lower bounds, respectively.
  • COAC provides significant performance gains over DRAC through centralized management and opportunistic accessing.
  • Delay-aware scheduling commonly uses equivalent rate constraints, Lyapunov optimization, or Markov decision processes.

V. CHALLENGING WORK AND OPEN ISSUES

The paper identifies edge caching, SDN, and NFV as unresolved issues for F-RAN deployment. It describes caching trade-offs and benefits while highlighting tensions between SDN centralization and F-RAN distribution.

  • F-RAN still faces open issues involving edge caching, software-defined networking, and network function virtualization.
  • A. Edge Caching: Edge caching trades transmission rate against storage, while local traffic storage can reduce constrained fronthaul burden and improve CRSP and CRRM performance.
  • A. Edge Caching: Reported caching benefits include alleviating fronthaul, backhaul, and backbone burden, reducing content-delivery latency, and enabling content-aware techniques.
  • A. Edge Caching: Small F-AP and F-UE caches often yield low-to-moderate hit ratios, motivating intelligent resource allocation and cooperative caching policies.
  • A. Edge Caching: Caching performance depends on jointly optimizing cache load, hit ratio, requests, hardware cost, and radio-resource cost, together with cache-release policies.
  • B. Software-Defined Networking: SDN can extend control toward the physical layer when CRSP and CRRM are incorporated into edge devices, supporting flexible and efficient network control.
  • B. Software-Defined Networking: Applying SDN in F-RANs remains challenging because SDN is centralized while F-RAN edge operation is distributed, and timely device reporting can burden fronthaul.

C. Network Function Virtualization

NFV separates network functions from proprietary hardware and supports programmable, migratable VNF connectivity in F-RANs. However, virtualizing SDN control and integrating NFV with F-RAN requirements remain open challenges.

  • C. Network Function Virtualization: NFV decouples network functions from dedicated hardware and transfers them to software-based applications.This virtualized approach aims to simplify and enhance telecommunications environments.
  • C. Network Function Virtualization: SDN for F-RANs enables programmable connectivity between VNFs managed by a VNF orchestrator.The orchestrator mimics the SDN controller's role.
  • C. Network Function Virtualization: NFV can virtualize the SDN controller on a cloud server and migrate it according to network needs.This provides location flexibility for control functions.
  • C. Network Function Virtualization: Virtualizing the SDN controller in F-RANs remains indistinct because of edge-device distribution characteristics.Related challenges include security, computing performance, VNF interconnection, portability, compatibility with legacy RANs, and management.
  • C. Network Function Virtualization: The article identifies edge caching, SDN, and NFV as areas requiring greater attention as F-RAN develops.The field is still relatively immature and contains outstanding problems requiring further investigation.
Loading 1506.04233v1…