PaperlayerView exact revision on arXiv

Source-linked AI summary

Base Station ON-OFF Switching in 5G Wireless Networks: Approaches and Challenges

Mingjie Feng, Shiwen Mao, Tao Jiang

arXiv:1703.09875v2cs.NI

TL;DR

5G densification is needed for much higher data rates but raises aggregate energy consumption, motivating BS ON-OFF switching. This paper analyzes switching challenges, reviews mechanisms across 5G scenarios, and discusses open problems; its traffic-aware example shows that adaptive sleep-state assignment can outperform single-state schemes on the energy-delay tradeoff.

  • Problem

    5G’s dense and heterogeneous deployments increase aggregate energy use while making BS ON-OFF switching harder because new technologies, architectures, and QoS requirements must be accommodated.

  • Method

    The paper analyzes technical aspects and challenges, reviews switching mechanisms across 5G network scenarios, and presents an abstract traffic-aware scheduling example.

  • Results

    The adaptive partial scheme outperforms single-sleep-state benchmark schemes on the energy-delay tradeoff in the indicative example.

  • Takeaways & Limitations

    Assigning BS sleep states according to traffic type can improve the balance between energy consumption and service delay.

Abstract

from arXiv · show

To achieve the expected 1000x data rates under the exponential growth of traffic demand, a large number of base stations (BS) or access points (AP) will be deployed in the fifth generation (5G) wireless systems, to support high data rate services and to provide seamless coverage. Although such BSs are expected to be small-scale with lower power, the aggregated energy consumption of all BSs would be remarkable, resulting in increased environmental and economic concerns. In existing cellular networks, turning off the under-utilized BSs is an efficient approach to conserve energy while preserving the quality of service (QoS) of mobile users. However, in 5G systems with new physical layer techniques and the highly heterogeneous network architecture, new challenges arise in the design of BS ON-OFF switching strategies. In this article, we begin with a discussion on the inherent technical challenges of BS ON-OFF switching. We then provide a comprehensive review of recent advances on switching mechanisms in different application scenarios. Finally, we present open research problems and conclude the paper.

I. INTRODUCTION

5G network densification supports much higher data rates and coverage but substantially increases aggregate energy consumption. BS ON-OFF switching can save energy, yet 5G technologies and architectures create new coverage, QoS, coordination, and scalability challenges.

  • 5G densification deploys many small-scale BSs or APs to support ubiquitous access and expected 1000x data rates.
  • More than 200 microcells can consume over 900 watts to provide coverage comparable to a 1500 m LTE macrocell.A typical LTE microcell consumes 25–40 watts depending on traffic load.
  • BS ON-OFF switching dynamically turns off under-utilized BSs as traffic fluctuates across time and space, improving energy efficiency.China Mobile estimated 36 million kWh of annual energy-consumption reduction from BS sleep control.
  • Interoperability with New Technologies: New 5G technologies require switching strategies to account for D2D-assisted coverage and coexistence with WiFi on unlicensed bands.The ON-OFF pattern may need adjustment to maintain WiFi users’ QoS.
  • Applications in New Network Architectures: C-RAN and mmWave architectures complicate switching because BBU shutdowns affect linked RRUs, while mmWave coverage is difficult to compensate after BS shutdown.Existing scheduling may not guarantee user QoS in these settings.
  • Large Number of BSs: Dense deployments create interdependent traffic, QoS, interference, and coordination decisions, making ON-OFF optimization combinatorial and difficult to scale.The challenge is amplified when CoMP allows users to be served by multiple BSs.

II. TECHNICAL ASPECTS AND CHALLENGES OF BS ON-OFF SWITCHING

BS ON-OFF switching research models energy use, traffic heterogeneity, sleep states, and wake-up decisions while balancing energy savings against delay and QoS. Traffic-aware adaptive sleep scheduling can improve the energy-delay tradeoff over single-state benchmarks.

  • BS energy models include static operating power and, in more detailed formulations, load-dependent power tied to users or active antennas.
  • Sleep-mode models increasingly retain nonzero detection power and distinguish multiple sleep levels because complete shutdown may prevent timely activation.Sleep power can be especially relevant for small-scale BSs without on-site cooling.
  • Traffic models must represent heterogeneous demands, including delay-sensitive and rate-sensitive services, rather than relying on a single traffic model.
  • B. Traffic Model and Traffic-Aware Scheduling: An adaptive partial scheme assigns standby and deep-sleep BSs to delay-sensitive and rate-sensitive traffic, respectively.The example uses half of the BSs in each sleep state when traffic is evenly split between the two demand types.
  • B. Traffic Model and Traffic-Aware Scheduling: The adaptive scheme outperforms two single-sleep-state benchmarks across energy-delay pairs obtained by varying wake-up intervals.Standby reduces energy through longer sleep, whereas deep sleep reduces delay through more frequent wake-ups.

C. Practical and Implementation Concerns

BS sleep control must coordinate QoS-sensitive switching with faster-changing user, channel, and traffic conditions. Periodic wake-up offers a tunable way to acquire system information, but its best interval depends on sleep mode and mobility.

  • Timescale of Operation: ON-OFF switching must be coordinated with user association and traffic offloading because BS states directly determine user QoS.User association changes more frequently than BS switching, while rapidly varying channels motivate time-averaged CSI.
  • How to Acquire System Information and Wake Up When Sleeping?: A sleeping BS can obtain information through neighboring BS assistance, periodic self-wake-up, or auxiliary devices that count tasks or capture UE signals.Periodic self-wake-up returns the BS to sleep when its activation criterion is not met.
  • How to Acquire System Information and Wake Up When Sleeping?: A proper wake-up interval maximizes EE across sleep modes and mobility scenarios.The example uses an always-active MBS, dynamically switched SBSs, standby and deep-sleep modes, and high- versus low-mobility conditions.
  • How to Acquire System Information and Wake Up When Sleeping?: Standby mode is better at small wake-up intervals, whereas deep sleep becomes more energy-efficient when intervals are large.Standby tracks traffic dynamics more closely but consumes more energy when load information is acquired infrequently.
  • How to Acquire System Information and Wake Up When Sleeping?: High mobility favors a shorter wake-up interval because small-cell user counts vary more rapidly.The figure compares EE against wake-up interval under different sleep modes and mobility levels.

D. Tradeoff with Other Performance Metrics

Switching off BSs saves energy but can increase handovers, delay, neighboring loads, and congestion. The literature therefore incorporates QoS constraints or metrics that explicitly balance energy against performance.

  • Tradeoff with Other Performance Metrics: BS switching can degrade user QoS through handovers, delayed wake-up, and increased traffic loads at neighboring BSs.In a two-tier HetNet, switching off many SBSs can overload the MBS and cause delay or congestion.
  • Tradeoff with Other Performance Metrics: QoS is handled either by adding requirements as formulation constraints or by using performance metrics that represent energy-performance tradeoffs.Examples include EE, defined as bits/Joule, and a weighted sum of energy consumption and delay.

E. Low Complexity Algorithms

BS ON-OFF scheduling is combinatorial and often NP-hard, especially when jointly optimized with multiple metrics. Reviewed solutions reduce complexity through decomposition, greedy procedures, transformations, distributed schemes, and dynamic programming.

  • Low Complexity Algorithms: BS ON-OFF scheduling is generally NP-hard and becomes mixed-integer programming when jointly optimized with multiple performance metrics.The difficulty grows with many BS variables and coupled design tasks.
  • Low Complexity Algorithms: Decomposition and iterative solution methods decouple BS switching from related tasks, while step-by-step greedy algorithms provide low-complexity alternatives.These approaches are used across several reviewed formulations.
  • Low Complexity Algorithms: Near-optimal solutions can be obtained by replacing a non-convex objective with a quasi-convex lower bound and maximizing that surrogate.The lower-bound formulation is presented as a tractable alternative.
  • Low Complexity Algorithms: Distributed schemes can reduce complexity by formulating interactions between users and BSs as a bidding game that determines user association.The initial game outcome provides the association strategy before subsequent decisions.
  • Low Complexity Algorithms: Renewable-energy settings jointly optimize energy allocation, BS scheduling, and resource allocation using a two-stage dynamic program with low complexity.The method targets minimum on-grid energy subject to average user blocking probability, and a HetNet extension considers energy-harvesting SBSs.

B. Massive MIMO HetNet

Massive MIMO, C-RAN, CoMP, and smart-grid settings alter the switching problem through channel-estimation overhead, functional decomposition, cooperation, and power-source uncertainty. These scenarios require switching decisions to account for coupled mappings, overhead, reliability, and energy supply.

  • Massive MIMO HetNet: In massive MIMO HetNets, increased MBS traffic requires more pilot symbols, reducing the proportion available for data and lowering average user throughput.Switching must consider the tradeoff between traffic load and channel-estimation overhead.
  • Cloud Radio Access Network: C-RAN separates processing in centralized BBUs from transmission and reception in distributed RRUs, making BBU switching more consequential than RRU-only switching.BBUs consume more power, but switching one affects all linked RRUs and their users.
  • Cloud Radio Access Network: C-RAN BBU scheduling couples ON-OFF states with BBU-RRU and RRU-user mappings, motivating joint fronthaul allocation, mapping, and scheduling.A load-balancing process activates another BBU if overload remains.
  • Coordinated Multipoint: CoMP switching requires physical-layer signal analysis because users may be served by multiple cooperating BSs and coordination increases overhead.Related work jointly considers BS and antenna switching with power allocation to maximize EE.
  • Smart Grid Empowered Wireless System: Smart-grid-powered wireless systems face uncertainty in renewable generation, power prices, and wireless conditions.Combining renewable and grid power is presented as a way to reduce operating cost and greenhouse-gas emissions.

F. Infrastructure Sharing Among Different Operators

Infrastructure sharing lets one operator switch off BSs and migrate traffic to another operator’s active BSs, but cooperation requires incentive mechanisms. D2D-assisted switching can save more energy while introducing UE energy and coordination constraints.

  • Operators can migrate traffic from switched-off BSs to co-located active BSs of another operator.
  • Infrastructure sharing requires incentive mechanisms to motivate cooperation between operators.
  • D2D communication permits BS shutdown despite coverage holes, but increases UE energy consumption.
  • D2D-assisted switching must jointly consider UE participation, path selection, link scheduling, and resource allocation.

2) MmWave Networks:

MmWave and heterogeneous networks complicate BS switching because blockage, beamforming, backhaul, handover, and MEC constraints affect coverage and service continuity. Effective scheduling must coordinate these coupled network resources.

  • MmWave Networks: MmWave blockage and propagation loss make switched-off BS coverage difficult to compensate through other BSs.
  • MmWave Networks: D2D communication can enhance indoor and outdoor coverage in blocked mmWave networks, but requires neighbor discovery, beamforming, and interference management.
  • Heterogeneous Networks with Wireless Backhaul: Wireless backhaul rates must support neighboring SBS traffic after switching, while also carrying timely scheduling information.
  • Heterogeneous Networks with Wireless Backhaul: Switching an SBS changes UE handovers, throughput, spectrum allocation, and channel-estimation overhead across active WBs and MUEs.
  • Mobile Edge Computing: MEC-enabled switching must account for task properties because edge storage and computation are limited relative to cloud servers.

4) Mobile Edge Computing:

Caching stations can improve QoS by storing frequently requested data, but their switching decisions must track changing content demand and the tasks stored at each station.

  • Caching stations store frequently requested data-intensive tasks in advance to provide high QoS.
  • Switching off a caching station affects every task stored there, requiring consideration beyond ordinary cellular traffic load.
  • Demand prediction from statistical information or pattern recognition can improve caching-station switching decisions.

6) M2M in IoT:

M2M relaying can enable AP or BS shutdown, but devices’ relay capabilities and update requirements constrain switching. Unlicensed-band coexistence and centralized control add further coordination and complexity challenges.

  • M2M in IoT: M2M direct transmission can let machines relay signals and create opportunities to turn off APs or BSs.
  • M2M in IoT: M2M relaying requires additional functionality on relay devices, limiting its applicability across heterogeneous IoT devices.
  • M2M in IoT: APs must be activated to balance service provision with storage and update-frequency requirements across machines.
  • Operation in Unlicensed Band: On unlicensed bands, BS shutdown can support Wi-Fi QoS but creates a tradeoff between 5G and existing-system users.
  • Operation in Unlicensed Band: LTE-U switching must coordinate user association and resource allocation while reducing centralized processing overhead and complexity.
  • Operation in Unlicensed Band: Clustering BSs with separate controllers is proposed to keep switching feasible when centralized network control becomes prohibitive.

1) Scalability:

The paper highlights machine learning as a potential way to improve scalable BS ON-OFF switching, while arguing that user satisfaction requires broader QoE modeling than conventional QoS metrics. It frames these issues within a review of 5G switching challenges, advances, and open problems.

  • 2) Application of Machine Learning Techniques:: Machine learning could predict user distribution and traffic loads so under-utilized BSs remain asleep while wake-up timing avoids additional delay.The paper identifies traffic prediction from historical data and user movement as a possible basis for switching decisions, but notes that implementation requires further study.
  • 1) Scalability:: Existing models mostly use QoS metrics, but user satisfaction also depends on outage, delay, throughput, congestion, video characteristics, and device energy consumption.Because users evaluate these factors differently, the paper argues for incorporating quality of experience into system design.
  • 1) Scalability:: The article identifies 5G BS ON-OFF switching challenges, reviews advances across wireless-network scenarios, and discusses open problems and potential solutions.Its scope combines technical analysis with a survey of switching mechanisms and an outlook for future research.
Loading 1703.09875v2…