Source-linked AI summary
System Architecture and Key Technologies for 5G Heterogeneous Cloud Radio Access Networks
Mugen Peng, Yong Li, Zhongyuan Zhao, Chonggang Wang
TL;DR
The paper addresses co-channel interference in heterogeneous radio access networks while pursuing higher coverage and capacity. It surveys H-CRAN architecture and key technologies, reporting gains reaching 40∼100% in weak coverage areas while identifying unresolved implementation challenges.
Problem
Co-channel interferences remain a problem in heterogeneous radio access networks, motivating large-scale cooperative processing.
Method
The article surveys H-CRAN architecture and key technologies, including advanced spatial signal processing, cooperative radio resource management, network function virtualization, and self-organization.
Results
40∼100% gain can be reached in weak coverage areas.
Takeaways & Limitations
H-CRAN research combines heterogeneous networks and cloud radio access networks to support higher system capacity and high throughput.
Takeaways & Limitations
A key problem for implementing the energy-harvesting approach remains unaddressed.
Abstract
from arXiv · showhide
Compared with the fourth generation (4G) cellular systems, the fifth generation wireless communication systems (5G) are anticipated to provide spectral and energy efficiency growth by a factor of at least 10, and the area throughput growth by a factor of at least 25. To achieve these goals, a heterogeneous cloud radio access network (H-CRAN) is presented in this article as the advanced wireless access network paradigm, where cloud computing is used to fulfill the centralized large-scale cooperative processing for suppressing co-channel interferences. The state-of-the-art research achievements in aspects of system architecture and key technologies for H-CRANs are surveyed. Particularly, Node C as a new communication entity is defined to converge the existing ancestral base stations and act as the base band unit (BBU) pool to manage all accessed remote radio heads (RRHs), and the software-defined H-CRAN system architecture is presented to be compatible with software-defined networks (SDN). The principles, performance gains and open issues of key technologies including adaptive large-scale cooperative spatial signal processing, cooperative radio resource management, network function virtualization, and self-organization are summarized. The major challenges in terms of fronthaul constrained resource allocation optimization and energy harvesting that may affect the promotion of H-CRANs are discussed as well.
I. INTRODUCTION
The paper motivates 5G H-CRANs by rising mobile traffic, stringent efficiency targets, and limitations of existing cellular, C-RAN, and HetNet architectures. It frames cloud-based centralized processing and heterogeneous integration as responses to these demands.
- Existing cellular networks are considered insufficient for traffic growth and energy efficiency because base-station power both overcomes path loss and causes interference.
- 5G is anticipated to deliver about 1000 times higher wireless area capacity and up to 90% lower energy consumption per service than 4G.
- H-CRANs combine heterogeneous networks with cloud computing to provide on-demand processing, storage, and network capacity.
- C-RAN centralizes baseband processing and enables cooperative interference mitigation, but fronthaul constraints and implementation complexity limit its scale.
B. 5G HetNet Solution
The paper presents H-CRANs as a 5G solution combining HetNets and C-RANs to support large-scale cooperative processing and networking. It surveys the associated technologies and identifies resource-allocation and energy-harvesting challenges.
- H-CRANs are described as backward compatible with different C-RAN and HetNet architectures while targeting improved spectral and energy efficiency.
- H-CRANs embed cloud computing into HetNets to realize large-scale cooperative signal processing and networking functionalities.
- The architecture decouples control and user planes, with MBSs providing control signaling and seamless coverage while RRHs provide hotspot data transmission.
- The surveyed technology areas include advanced spatial signal processing, cooperative radio resource management, NFV, and self-organizing networks.
- Fr onthault-constrained resource allocation optimization and energy harvesting are identified as challenging issues for H-CRAN promotion.
II. APPLICATION ARCHITECTURE AND SYSTEM COMPONENTS
The H-CRAN application architecture uses heterogeneous access entities, dense RRHs, cloud computing, and Node C to provide flexible, scalable access and adaptive resource operation. Its software-defined design supports SDN compatibility and traffic-responsive coordination.
- Node C converges existing communication entities and manages RRHs as a BBU pool with centralized cooperative processing capabilities.
- RRHs provide high-speed hotspot transmission, while ACEs deliver control overhead and cell-specific reference signals for broader coverage.
- Universal plug-and-play access allows newly deployed RRHs and existing ACEs to connect to Node C automatically.
- Hundreds of RRHs and several tens of ACEs can be served simultaneously, enabling efficient control and user-plane separation.
- The software-defined architecture supports SDN-compatible signaling and can reduce radio connection and release overhead by moving beyond purely connection-oriented operation.
- Traffic-adaptive operation puts underused RRHs to sleep, activates heterogeneous access jointly in high-load zones, and lets heavily loaded RRHs borrow neighboring resources.
B. System Components of H-CRANs
H-CRANs organize operation across user, control, and management planes within a software-defined architecture. Centralized programmability supports network-wide decisions, automation, adaptive processing, and self-organization.
- Planes: The user plane carries traffic, the control plane manages signaling and resource allocation, and the management plane administers operations and plane interactions.
- Planes: RRHs are configured with the control plane, whereas ACEs incorporate both control and user planes.
- Software-defined architecture: The software-defined H-CRAN delivers SDN control information to Node C through the standardized OpenFlow southbound interface.
- Software-defined architecture: A logically centralized controller facilitates global decision-making, automated configuration, and network-wide traffic-forwarding decisions.
- Key functionalities: Key functionalities include self-organization, radio-resource cloudization, cognitive processing, and big-data-supported intelligent networking.
- Key functionalities: Cognitive processing coordinates RRHs and ACEs under lower load, while resource cloudization improves radio-resource reuse under high load.
III. PROMISING KEY TECHNOLOGIES
The paper presents LS-CSSP as a cooperative interference-cancellation technology for software-defined H-CRANs, alongside LS-CRRM, NFV, and SON. Centralized and distributed LS-CSSP offer distinct capacity, coverage, fronthaul, and interference-management benefits but face CSI, complexity, and scaling challenges.
- Promising key technologies: LS-CSSP uses cooperation across H-CRAN nodes to cancel interference through software-defined architectures.The paper identifies LS-CSSP as a key functionality for exploiting H-CRANs and describes it as collaborative interference cancellation.
- Distributed LS-CSSP: Distributed LS-CSSP can strengthen uplink signals and suppress inter-RRH interference, with 40 ∼100% gains in weak coverage areas and gains exceeding 100% in some areas.Weak coverage is defined as RSRP lower than -95dBm.
- Open issues: Distributed LS-CSSP must balance spectral-efficiency gains against CSI signaling, accuracy, instantaneity, fronthaul capacity, and implementation complexity.Small clusters may limit improvement, whereas excessive participation makes signaling intractable and degrades spectral efficiency.
- Centralized LS-CSSP: Centralized LS-CSSP uses many co-located ACE antennas to improve capacity and coverage while reducing antenna deployment complexity.The centralized configuration is described as using hundreds of low-power antennas at a co-located ACE site.
- Centralized LS-CSSP: 10 times or more capacity increase and radiated EE improvement on the order of 100 times are reported for centralized LS-CSSP.These gains are attributed to the centralized processing configuration.
B. Large-Scale Cooperative Radio Resource Managements (LS-CRRM)
LS-CRRM coordinates association, radio blocks, power, admission, and queue information to manage interference and traffic in H-CRANs. Enhanced S-FFR suppresses inter-tier interference, while JARP improves utility and queue performance relative to the stated baselines.
- Enhanced S-FFR: Enhanced S-FFR suppresses inter-tier interference between ACEs and RRHs by allocating resources according to RUE QoS requirements.Only QoS requirements need to be distinguished for RUEs in the enhanced scheme.
- Enhanced S-FFR: Enhanced S-FFR allows low-QoS RUEs and HUEs to share radio resources, unlike traditional S-FFR's orthogonal allocation for cell-edge UEs.The traditional scheme is reported to reduce spectral efficiency significantly in H-CRANs.
- Enhanced S-FFR: The enhanced scheme divides resources into Ω1 for high-rate-constrained RUEs and Ω2 for low-rate-constrained RUEs and HUEs.Centralized processing enables inter-RRH interference avoidance and radio-resource reuse among RRHs.
- LS-CRRM objectives: LS-CRRM jointly considers traffic admission, RRH/MBS association, RB allocation, and power allocation under power-consumption constraints.Queue state information is included to support fairness and radio-resource utilization.
- LS-CRRM evaluation: As RUE traffic arrival rates increase, HUE average queue length grows and system throughput utility increases with a diminishing slope.At higher arrival rates, some RUE traffic must be denied to stabilize queues.
- LS-CRRM evaluation: JARP achieves higher utility and smaller average queue length than JAREP and RAJRP under all traffic arrival rates.The comparison is based on joint association, RB, and power allocation against the two stated baselines.
C. Network function virtualization (NFV) in H-CRANs
NFV virtualizes H-CRAN radio and computing resources through software-defined infrastructure, supporting flexible orchestration, isolation, and control. The approach offers potential service-delivery and cost benefits but must address competing slice isolation, utilization, signaling, and retransmission requirements.
- NFV and SDN: NFV in SDNs can support packet processing in the user plane and control functions in the control plane.The paper associates this integration with potential improvements in service delivery and reductions in overall costs.
- NFV architecture: NFV cloudizes physical radio resources and virtualizes computing resources across intra-RANs and inter-RANs in H-CRANs.The virtualized infrastructure is managed through orchestration and resource assignment mechanisms.
- NFV architecture: Node C supports rapid VM creation and deletion, customizable distributions, kernel and driver customization, and administration.The infrastructure uses high-volume computing servers, storage devices, and network switches organized by orchestration.
- Isolation and transparency: NFV virtualizes heterogeneous ACEs and RRHs while keeping virtual nodes and virtual-layer exchanges transparent for isolation and privacy.Minimal primitives may expose computational-resource usage and traffic consumption to resource controllers.
- NFV design issues: H-CRAN NFV must support coexistence of bandwidth-based and resource-based slice reservations because achieved slice bandwidth varies with channel quality and scheduling.This variability complicates predictable resource provisioning across slices.
- NFV design issues: Uplink traffic randomness makes isolation and efficient resource utilization across slices conflicting goals, while signaling and retransmissions add overhead.These overheads hurt resource utilization and must be accounted for in H-CRAN resource sharing.
D. Large-Scale Self-Organizing H-CRANs (LS-SON)
Large-scale self-organizing H-CRANs centralize network-management functions in Node C while coordinating heterogeneous RAN resources through cloud computing. LS-SON integrates automated planning, configuration, optimization, and healing to improve operational efficiency and support energy and interference management.
- Performance: LS-SON improves operational efficiency, suppresses co-channel interference, and supports improvements in both energy efficiency and spectral efficiency.The architecture also enables radio-resource sharing among RRHs and can reduce operational costs.
- Architecture: LS-SON integrates ultra learning, planning, configuration, and optimization into one largely automatic process with minimal manual intervention.Cloud centralization reduces management complexity in heterogeneous, virtualized H-CRANs.
- Architecture: Node C centralizes inter-RAN and intra-RAN SON functions for converged RANs and cooperatively processed RRHs.This architecture implements self-configuration, self-optimization, and self-healing in centralized mode.
- Self-configuration: RRHs need intelligent radio-resource self-assignment, whereas ACEs additionally require self-configuration of physical cell identifiers and radio resources.Node C handles ACE self-configuration centrally.
- Self-optimization: Energy saving, mobility load balancing, and mobility robustness optimization are key self-optimization cases in LS-SON.Dynamic energy saving switches off unneeded RRHs and ACEs while maintaining desired QoS; mobility load balancing is centralized in Node C.
- Self-healing: Self-healing detects or predicts problems, identifies their root causes, and applies corrective actions to restore service.Corrective actions may be full or partial, definite or temporary.
IV. CHALLENGES AND OPEN ISSUES IN H-CRANS
Although H-CRAN techniques have achieved initial progress, important challenges remain for their practical development. The article identifies fronthaul-constrained resource allocation, energy harvesting, and backward-compatible standardization as open issues.
- Open issues: H-CRANs still face unresolved challenges in optimal resource allocation over constrained fronthaul, energy harvesting, and backward-compatible standard development.These issues are presented as challenges that may affect further development of H-CRANs.
A. Fronthaul Constraints and Performance Optimization
Fronthaul capacity and latency constrain H-CRAN throughput and cooperative processing, while practical solutions may not meet QoS requirements everywhere. The article calls for lower-complexity optimization methods that account for non-ideal fronthaul conditions.
- Impact: Fronthaul characteristics affect UE throughput and overall H-CRAN performance by constraining data transport and cooperative processing.High bit rate and low latency enable larger-scale cooperation and more efficient use of scarce resources.
- Impact: Practical fronthaul solutions may lack sufficient end-to-end performance to meet desired QoS requirements everywhere.Limited capacity and time delay are identified as inevitable practical constraints.
- Optimization: Resource allocation and admission control under non-ideal fronthaul constraints form an NP-complete optimization problem.In moderately dense RRH deployments, computational and memory costs may prevent optimal solutions.
- Optimization: Flexible iterative algorithms with limited complexity are needed to improve H-CRAN spectral- and energy-efficiency performance under fronthaul constraints.The article identifies constrained optimization as a direction for future work.
B. Energy Harvesting in H-CRANs
Energy harvesting introduces time- and space-varying renewable supply into H-CRAN resource management. The article highlights adaptive transmission power and Node C-assisted energy cooperation, while identifying node incentives as unresolved.
- Constraints: Energy storage based on capacity-limited, expensive batteries is insufficient to manage renewable-energy fluctuations in 5G systems.The article frames these fluctuations and shortages as a challenge for ultra-dense RRHs and ACEs.
- Energy cooperation: Node C can support energy cooperation by sharing excess energy among RRHs and ACEs, but motivating large-scale participating nodes remains unaddressed.The unresolved issue concerns cooperation among ultra-dense nodes using renewable energy.
- Adaptive management: Energy harvesting should adapt to changing spectrum requirements and traffic volume, including peak-hour transmission demands.RRHs and ACEs may increase transmission power to improve spectral efficiency when traffic is high.
- Energy variability: Renewable energy availability varies over time and space, making fixed-transmission-power assumptions inadequate for H-CRAN SE and EE optimization.Solar and wind sources are specifically identified as intermittent supplies.
- Paper scope: The article surveys H-CRAN architectures and technologies, including spatial signal processing, cooperative resource management, NFV, and self-organizing networks.It also discusses fronthaul constraints, performance optimization, and energy harvesting as open issues.