PaperlayerView exact revision on arXiv

Source-linked AI summary

A Survey on Mobile Edge Networks: Convergence of Computing, Caching and Communications

Shuo Wang, Xing Zhang, Yan Zhang, Lin Wang, Juwo Yang, Wenbo Wang

arXiv:1703.10750v1cs.NI

TL;DR

Explosive traffic growth and centralized-network latency and backhaul constraints motivate architectures that move computing, caching, and network functions to the edge. This survey synthesizes mobile edge architectures and research on computing, caching, communications, applications, and enablers, reporting benefits including lower latency, energy savings, and reduced transmission cost. It also identifies mobility, real-time resource management, and open research challenges.

  • Problem

    Centralized mobile networks face heavy backhaul burdens and long latency as smart-device traffic and application demands grow.

  • Method

    The paper comprehensively surveys mobile edge architectures and research on computing, caching, communications, applications, enabling technologies, and future challenges.

  • Results

    The survey reports reduced latency and energy consumption from edge offloading, up to 88% lower cost with multicast-aware caching, and up to 42% lower mobile-device energy consumption through cloudlet offloading.

  • Takeaways & Limitations

    Mobile edge networks integrate computing, caching, and communication resources near users to support latency-sensitive delivery, offloading, and real-time analytics.

Abstract

from arXiv · show

As the explosive growth of smart devices and the advent of many new applications, traffic volume has been growing exponentially. The traditional centralized network architecture cannot accommodate such user demands due to heavy burden on the backhaul links and long latency. Therefore, new architectures which bring network functions and contents to the network edge are proposed, i.e., mobile edge computing and caching. Mobile edge networks provide cloud computing and caching capabilities at the edge of cellular networks. In this survey, we make an exhaustive review on the state-of-the-art research efforts on mobile edge networks. We first give an overview of mobile edge networks including definition, architecture and advantages. Next, a comprehensive survey of issues on computing, caching and communication techniques at the network edge is presented respectively. The applications and use cases of mobile edge networks are discussed. Subsequently, the key enablers of mobile edge networks such as cloud technology, SDN/NFV and smart devices are discussed. Finally, open research challenges and future directions are presented as well.

I. INTRODUCTION

Mobile edge networks address explosive mobile traffic and emerging application demands by moving computing, caching, and network functions closer to users. The paper surveys their architectures, technologies, applications, enablers, and research challenges.

  • Mobile data traffic grew 4000-fold over the past decade, while new applications and devices impose stricter data-rate and latency requirements.
  • Mobile cloud computing suffers from long latency and backhaul bandwidth limitations because mobile devices reach distant Internet clouds.
  • Mobile edge computing places cloud servers at base stations, providing proximity, low latency, high bandwidth, real-time radio information, and location awareness.
  • Fog computing, cloudlets, and edge caching extend edge-based processing or content delivery toward users, serving IoT, multimedia, and other applications.
  • The survey covers mobile edge architectures, computing and caching techniques, applications, enabling technologies, and open research challenges.

II. OVERVIEW OF MOBILE EDGE NETWORKS

Mobile edge networks move flexible computing, storage, caching, and communication resources toward users through architectures such as MEC, fog computing, cloudlets, and edge caching. These paradigms reduce distance-related costs and support proximity-sensitive services, while introducing deployment and mobility-management challenges.

  • B. Architecture of Mobile Edge Networks: MEC uses virtualized platforms and servers near base stations, with deployment locations selected according to latency, resources, availability, scalability, and cost.
  • C. Edge Caching: Proactive edge caching exploits spatial and temporal traffic variation to store popular content during off-peak periods and can use social structures with D2D dissemination.
  • A. What is Mobile Edge Networks: Mobile edge networks deploy computing and storage resources across the radio access network, edge routers, gateways, and mobile devices using SDN and NFV.
  • B. Architecture of Mobile Edge Networks: Fog computing targets IoT through massively distributed fog nodes that collaborate with end-user or near-user devices for processing and storage.
  • B. Architecture of Mobile Edge Networks: Cloudlets form a device–cloudlet–cloud architecture, support near-real-time application provisioning and virtual-machine handoff, and can operate at Wi-Fi or LTE access points.

3) Cloudlet:

Mobile edge networks place computing, storage, and communication resources closer to users than centralized architectures. This proximity can reduce latency, operating costs, energy consumption, and traffic load.

  • 3) Cloudlet:: Edge computing can improve response times by 51% compared with cloud offloading for Wi-Fi and 4G LTE networks.The cited experiments concern cloudlet offloading.
  • 3) Cloudlet:: Edge-server deployment can save operating costs by up to 67% for bandwidth-hungry and compute-intensive applications.
  • 3) Cloudlet:: Proactive caching can provide backhaul savings of up to 22%, with higher gains possible when storage capacity increases.
  • 3) Cloudlet:: Offloading computation to cloudlets can reduce mobile-device energy consumption by up to 42% compared with cloud offloading.
  • 3) Cloudlet:: Edge servers and D2D communication can provide proximity services and reduce traffic load on the radio access network.
  • 3) Cloudlet:: MEC servers can use network- and device-level context information to allocate resources more efficiently and support location-based applications.

5) Utilization of Context Information:

Computing at the mobile edge addresses constrained device computation and battery resources through task offloading. The surveyed objectives and approaches span energy, capacity, latency, scheduling, and multi-user coordination.

  • 5) Utilization of Context Information:: Edge computing enables compute-intensive tasks to be offloaded from resource-constrained mobile devices to more powerful edge servers.
  • 5) Utilization of Context Information:: Energy-aware offloading schemes jointly consider task computation and file transmission, with iterative channel allocation designed to minimize system energy cost.
  • 5) Utilization of Context Information:: 5G networks require 1000 times higher mobile data volume per area than current 4G LTE networks, making offloading one capacity-enhancing technology.
  • 5) Utilization of Context Information:: 1 ms RTT is the stated latency requirement for next-generation 5G networks, compared with 10 ms RTT in 4G.
  • 5) Utilization of Context Information:: Offloading decisions must account for scenarios including single users, multiple users, energy harvesting, wireless-channel variation, and vehicular networks.
  • 5) Utilization of Context Information:: The multi-user computation offloading problem is NP-hard, motivating distributed game-theoretic approaches and joint radio-resource and computational-resource optimization.

2) Multi-user Case:

Mobile edge networks support multiple deployment and resource-management strategies beyond centralized MEC-server offloading. The surveyed approaches include device cooperation, mobility awareness, fog-cloud coordination, and practical platform deployments.

  • 2) Multi-user Case:: IoT devices can offload memory objects to LTE eNodeB edge clouds, while video-call encoding can be offloaded to MEC servers to reduce latency and energy consumption.
  • 2) Multi-user Case:: Task-offloading schemes can jointly consider energy consumption, time delay, and execution-unit cost when designing optimal policies.
  • 2) Multi-user Case:: Co-located mobile devices can provide cloud services at the edge through D2D communication, requiring task scheduling distinct from server offloading.
  • 2) Multi-user Case:: User mobility changes contact time with MEC servers and therefore affects where and what computation tasks should be offloaded.
  • 2) Multi-user Case:: Fog-cloud cooperation is used to study workload allocation and tradeoffs between power consumption and delay.
  • 2) Multi-user Case:: MEC and fog platforms have been implemented and evaluated with macro-cell, D2D-based, and small-cell-based connections to LTE networks.

1) MBS Caching:

Mobile edge caching places content across cellular-network layers and user devices, with caching decisions shaped by popularity, replacement policy, and wireless-network conditions. The surveyed literature covers MBS, SBS, D2D, and dynamic-popularity approaches.

  • 1) MBS Caching:: Caching at macro base stations can improve video capacity and reduce video stalling probability when combined with backhaul and wireless-channel scheduling.
  • 1) MBS Caching:: Caching at small base stations benefits from their proximity to end users and typically higher data rates.
  • 1) MBS Caching:: D2D caching schemes exploit mobile-device storage and may incorporate users’ social relations and common interests.
  • 1) MBS Caching:: Content popularity is used to select what to cache and maximize cache-hit probability, but static IRM models do not represent time-varying popularity.
  • 1) MBS Caching:: LFU and LRU are simple and efficient for uniform-size objects, but they ignore download latency and object size; MPV can have low hit probability in limited RAN caches.

2) User Preference Based Policies:

Edge-caching research develops policies that use user preferences, learned popularity, cooperation, data characteristics, and mobility to place and replace content. These approaches address changing demand and wireless-network constraints while coordinating cache resources across locations.

  • User preference profiles capture cell-specific video-category interests because local popularity can differ substantially from national popularity.
  • Learning-based policies estimate time-varying popularity and apply reinforcement learning or Q-learning to distributed caching and cache replacement.
  • Cooperative caching coordinates cache locations to increase cache-served traffic and reduce bandwidth cost, using optimization and approximation methods.
  • Caching policies must distinguish multimedia files, which benefit from popular-content caching, from short-lived IoT data requiring freshness-aware decisions.
  • Mobility-aware caching models spatial and temporal movement or Markovian mobility to improve delivery across base stations.

F. Impact on System Performance

Edge caching improves capacity, delay, and energy-related performance, while cache-enabled network design also interacts with base-station density and high-mobility mm-wave communication. These results motivate joint consideration of caching, deployment, and communication technologies.

  • 1) Capacity: 3 times: caching in the RAN can improve system capacity compared with having no cache.
  • 2) Delay: Edge caching reduces content-delivery delay through cache proximity and can improve video QoE by jointly scheduling backhaul and wireless channels.
  • 1) Capacity: 3/4: cache-enabled helper density can be reduced relative to pico-BS density without caching while achieving the same area spectral efficiency.
  • 4) Energy Efficiency: Caching improves energy efficiency under a small file catalog, with multiple pico base stations more energy efficient than a macro base station.
  • A. mmWave Communication: Precaching at mm-wave base stations can provide consistent high-quality video streaming for high-mobility users facing short connections and frequent handoffs.

C. Transmission Schemes

Transmission schemes in mobile edge networks exploit multicast, caching, computing, and coordinated scheduling to reduce traffic costs, manage interference, and conserve constrained communication resources. Their design depends on application, mobility, and resource requirements.

  • C. Transmission Schemes: Up to 88%: multicast-aware transmission can reduce cost compared with caching schemes using unicast transmission.
  • D. Interference Management: Caching simplifies interference-network topology, making interference management easier while reducing backhaul load and CSI-feedback overhead and improving throughput.
  • E. Communication Resources Allocation and Scheduling: More than 50%: joint video-aware backhaul scheduling and caching can improve capacity over conventional policies.
  • F. Synergy of Communication, Computing and Caching: At least 50%: content slimming can reduce transmission bandwidth consumption versus H.264 without sacrificing video quality or visual experience.
  • F. Synergy of Communication, Computing and Caching: The optimal placement of computing and caching resources depends on application type, user mobility, and available communication resources.

D. Video Streaming and Analysis

Mobile edge networks support video analysis, healthcare, smart-grid, smart-home, smart-city, vehicle, and cognitive-assistance applications by processing data near users and devices. The surveyed use cases emphasize low latency, local context, and distributed analytics.

  • D. Video Streaming and Analysis: Edge caching avoids redundant video transport through the core network, while MEC enables video analysis at capable edge platforms.
  • D. Video Streaming and Analysis: Fog computing supports IoT edge analytics, including crowd-sourced video processing through VM-based cloudlets.
  • D. Video Streaming and Analysis: Healthcare experiments found fog-based systems responding faster and using less energy than cloud-only approaches.
  • D. Video Streaming and Analysis: Smart-grid edge computing combines fog-based local collection and processing with cloud-based global coverage and long-term data storage.
  • D. Video Streaming and Analysis: MEC supports localized, low-latency services for smart homes, smart cities, connected vehicles, V2X communication, and cognitive assistance.

H. Wireless Big Data Analysis

Mobile edge networks combine cloud, SDN/NFV, and device-side intelligence to support flexible, scalable, and efficient network operation. The survey identifies heterogeneity and stringent performance requirements as continuing research challenges.

  • Enabling technologies: Cloud technology extends computing capabilities to mobile-network edges through virtual machines deployed on general-purpose servers.The paper describes deployment at locations such as base stations and gateways.
  • Enabling technologies: SDN separates control and data planes, while NFV virtualizes network functions on general-purpose platforms to improve network programmability, scalability, and flexibility.NFV can also reduce operators’ capital and operating expenses.
  • Smarter devices: Device-to-device communication can reduce signaling resources, transmission latency, and energy use while improving spectral efficiency.The paper attributes these benefits to direct communication between nearby devices and lower path losses than base-station links.
  • Open challenges: The survey frames mobile edge networks as an architectural evolution with improved QoE performance and flexibility, alongside broad research opportunities.The cited application table is identified as covering mobile edge network applications and use cases.
  • Open challenges: Mobile edge networks must address heterogeneous networking, communication, and devices under a unified architecture.The paper links this heterogeneity with issues including asynchronism and non-orthogonality.

2) Computation Modeling:

The survey highlights computation modeling, dynamic resource management, mobility, pricing, scalability, and security as interconnected concerns for mobile edge networks. These concerns arise because edge resources and users are heterogeneous, mobile, and dynamically managed.

  • Computation modeling: Current analyses often model computation resources as CPU cycles per second, but more accurate models are needed to reflect computing characteristics.The paper presents this as a limitation affecting the validation of analytical and simulation results.
  • Resource management: Real-time applications require dynamic resource management that schedules analytic tasks to suitable edge servers while meeting latency and throughput requirements.The paper gives VR/AR and e-Health as examples of applications requiring real-time analytics.
  • Mobility: User mobility affects caching and computation-offloading decisions because frequent movement causes handovers among edge servers.The survey calls for mobility management supporting seamless access under horizontal and vertical mobility.
  • Pricing: Dynamic resource allocation creates pricing challenges because stakeholder profits must be balanced when users are price-sensitive.The survey notes game-theoretic work on pricing and resource allocation in video caching systems.
  • Scalability: Scaling services for growing numbers of mobile and IoT devices requires load balancing and flexible management of edge nodes.The paper identifies cloud orchestration as one possible means of providing network scalability.
  • Security and privacy: Edge deployment introduces security and privacy challenges because device information is exposed, links may be intermittent, and D2D cooperation can threaten user privacy.Existing cloud security solutions may not fit edge environments, while privacy-preserving mechanisms include encryption and trusted helper entities.

9) User Participation:

Mobile edge networks can use user-terminal resources, context information, and caching-aware access decisions, but participation incentives and integrated resource allocation remain open issues. The survey synthesizes these research directions across computing, caching, and communications.

  • User participation: Using available user-terminal resources requires incentive mechanisms because cooperation depends on users’ willingness to participate.The paper specifically connects incentives with computation-offloading and caching strategies.
  • Context and data: Wireless big data and context information can support network analysis, design, and popularity estimation in edge caching systems.The survey identifies user-information data as an example of data useful for estimating content popularity.
  • Caching: Caching research includes cache placement and content delivery, with more efficient online cache updating during delivery identified as a future direction.The paper distinguishes delivery-phase updating from the more extensively studied placement phase.
  • Context and data: Context-aware applications can use location, nearby-user, and environmental-resource information to allocate resources proactively across MEC platforms.The survey identifies application-, network-, and device-level context as usable information.
  • Caching: Cache-aware user association can connect users to base stations storing requested content, potentially improving QoE and alleviating backhaul limitations.This changes association decisions beyond nearest-distance or SINR-based selection.
  • Integration: Efficiently integrating computing, storage, and communication resources remains unresolved, motivating more comprehensive resource-allocation schemes.The survey presents integration as an ongoing research topic as networks evolve.
  • Survey scope: The survey organizes mobile edge research around a paradigm integrating computing, caching, and communication resources, and identifies applications, challenges, and future directions.It surveys architectures, offloading, caching taxonomies, communication synergies, applications, and use cases.
Loading 1703.10750v1…