PaperlayerView exact revision on arXiv

Source-linked AI summary

A Survey of Multi-Access Edge Computing in 5G and Beyond: Fundamentals, Technology Integration, and State-of-the-Art

Quoc-Viet Pham, Fang Fang, Vu Nguyen Ha, Md. Jalil Piran, Mai Le, Long Bao Le, Won-Joo Hwang, Zhiguo Ding

arXiv:1906.08452v2cs.NIcs.DCcs.IT

TL;DR

The paper addresses the challenge of supporting compute-intensive 5G and IoT applications despite constrained devices and rapidly increasing demands. It surveys MEC fundamentals, its integration with 5G and beyond technologies, and related evaluations and future directions. The survey concludes that these integrations support massive IoT, self-sufficient systems, improved network performance, expanded connectivity, and viable MEC service economics.

  • Problem

    Rapidly growing traffic and computation demands exceed the storage and processing capabilities of many mobile devices, while prior MEC surveys covered limited aspects or provided high-level 5G discussions.

  • Method

    The paper conducts a holistic survey of MEC fundamentals, use cases, integration with 5G and beyond technologies, testbeds, experimental evaluations, open-source activities, lessons learned, and future directions.

  • Results

    MEC integration with 5G technologies supports massive IoT, self-sustainability, improved network performance and adaptability, expanded terrestrial connectivity, and MEC service economics.

  • Takeaways & Limitations

    MEC is presented as an edge-based approach for supporting latency-sensitive applications and diverse 5G services through proximity, local processing, and technology integration.

  • Takeaways & Limitations

    The survey identifies unresolved challenges involving resource management, mobility, reliability, and service continuity in distributed MEC environments.

Abstract

from arXiv · show

Driven by the emergence of new compute-intensive applications and the vision of the Internet of Things (IoT), it is foreseen that the emerging 5G network will face an unprecedented increase in traffic volume and computation demands. However, end users mostly have limited storage capacities and finite processing capabilities, thus how to run compute-intensive applications on resource-constrained users has recently become a natural concern. Mobile edge computing (MEC), a key technology in the emerging fifth generation (5G) network, can optimize mobile resources by hosting compute-intensive applications, process large data before sending to the cloud, provide the cloud computing capabilities within the radio access network (RAN) in close proximity to mobile users, and offer context-aware services with the help of RAN information. Therefore, MEC enables a wide variety of applications, where the real-time response is strictly required, e.g., driverless vehicles, augmented reality, robotics, and immerse media. Indeed, the paradigm shift from 4G to 5G could become a reality with the advent of new technological concepts. The successful realization of MEC in the 5G network is still in its infancy and demands for constant efforts from both academic and industry communities. In this survey, we first provide a holistic overview of MEC technology and its potential use cases and applications. Then, we outline up-to-date researches on the integration of MEC with the new technologies that will be deployed in 5G and beyond. We also summarize testbeds and experimental evaluations, and open source activities, for edge computing. We further summarize lessons learned from state-of-the-art research works as well as discuss challenges and potential future directions for MEC research.

I. INTRODUCTION

This survey motivates MEC as a response to rising 5G traffic and computation demands, then positions its contribution as a comprehensive review of MEC fundamentals and integration with technologies beyond prior focused surveys.

  • Motivation: 5G traffic and application demands are growing rapidly while mobile devices retain limited storage and processing capabilities.Projected global traffic reached 122 EB per month by 2022, while AR/VR traffic was expected to increase twelvefold.
  • From Cloud to Edge: Mobile cloud computing moves computation and storage away from devices but suffers from user distribution and latency constraints.Cloudlets and fog computing extend cloud resources toward the edge, although they are not integrated into mobile network architecture.
  • MEC Concept: MEC places cloud-computing capabilities within the RAN near mobile users, enabling applications such as autonomous vehicles, VR/AR, robotics, and immersive media.The MEC concept was initiated by ETSI’s MEC Industry Specification Group in late 2014.
  • Prior Work: Earlier surveys generally focused on specific MEC aspects, including overviews, offloading, resource management, enabling technologies, or mathematical frameworks.The prior literature also covered architectures, deployment scenarios, testbeds, security, privacy, and service orchestration.
  • Contributions: This survey provides a holistic MEC overview and reviews its integration with forthcoming 5G and beyond technologies and scenarios.Covered areas include NOMA, wireless power transfer and energy harvesting, UAVs, IoT, heterogeneous C-RAN, and machine learning.

C. Paper Organization

The paper organizes its survey around MEC fundamentals, 5G integration, applications, testbeds, and future directions, while describing MEC’s characteristics, opportunities, and unresolved deployment challenges.

  • Paper Organization: The survey covers MEC fundamentals, 5G integration, use cases, testbeds, implementation, lessons learned, and potential future works.Its major technology-focused sections address NOMA, WPT and EH, UAV communications, IoT, heterogeneous C-RAN, and machine learning.
  • Fundamentals: MEC provides an IT service environment and cloud-computing capabilities within the RAN and close to mobile subscribers.Its demand is associated with smart and IoT devices, increasing data volume and velocity, and high-bandwidth, low-latency applications.
  • Market Drivers: MEC’s market drivers include technical integration, potential use cases, business transformation, and industry collaboration.The paper links deployment to cooperation among mobile operators, service providers, vendors, and users.
  • Fundamentals: MEC is characterized by on-premises operation, proximity, lower latency, location awareness, and network contextual information.These features support local resources, user-related analytics, latency-sensitive processing, location services, and QoS optimization.
  • Challenges: Distributed resource management is difficult because MEC has finite resources, growing application demands, heterogeneous servers, and potentially congested wireless backhaul.Mobility can also cause frequent handovers, service disruption, unavailable computation results, interference variation, and time-varying computing resources.

4) Coexistence of distributed MEC and centralized cloud:

Distributed MEC handles latency-critical computation near users, while centralized cloud resources remain useful for compute-intensive, delay-tolerant tasks. Their coexistence requires coordination across MEC architecture and 5G network functions.

  • 4) Coexistence of distributed MEC and centralized cloud: Distributed MEC reduces end-to-end delay for latency-critical computation by processing tasks at the network edge.Centralized cloud reliance can be problematic for latency-critical services because of traffic congestion and transmission delay.
  • 4) Coexistence of distributed MEC and centralized cloud: A hybrid strategy assigns latency-critical computations to distributed MEC servers and compute-intensive, delay-tolerant tasks to the cloud.
  • 5G integration: Integration with the 5G service-based architecture adds mechanisms for traffic steering, local routing, session continuity, network capability exposure, QoS, and charging.
  • 5G integration: The 5G architecture exposes network functions and policies that support MEC deployment, including mobility, sessions, slicing, service discovery, identification, authentication, and user-plane operations.
  • MEC architecture: The MEC reference architecture includes system-level orchestration and host-level resources, with the MEC orchestrator selecting hosts and managing application placement.The orchestrator maintains information about hosts, resources, services, and topology, and can trigger application instantiation or relocation.

C. MEC in 5G and Beyond

MEC in 5G and beyond spans consumer, operator, third-party, and network-performance applications. The survey examines its integration with radio, architectural, application, power-supply, and machine-learning technologies.

  • MEC in 5G and Beyond: MEC use cases fall into consumer-oriented services, operator and third-party services, and network performance and QoE improvements.Examples include V2X, big data, device location tracking, security, analytics, indoor positioning, and content pushing.
  • MEC in 5G and Beyond: MEC integration with forthcoming 5G technologies is presented as necessary to achieve added value in MEC systems.
  • Radio access technologies: NOMA, mmWave, and massive MIMO are reviewed for their potential to support massive connectivity, high data rates, low latency, and greater computing capability.
  • Survey scope: The survey covers H-CRAN, IoT, UAV, WPT and energy harvesting, and machine learning alongside MEC.Its scope includes radio access, network architectures and scenarios, applications, power supply, and performance improvement.

III. MEC WITH NON-ORTHOGONAL MULTIPLE ACCESS

NOMA and MEC are presented as complementary 5G technologies: NOMA supports multi-user access, while MEC places computing resources near users. Their integration may improve connectivity, latency, energy use, and system performance.

  • A. Fundamentals of NOMA: NOMA uses superposition coding at the base station and interference cancellation at users to let multiple users share the same time-frequency resources.
  • A. Fundamentals of NOMA: Compared with OMA, NOMA is associated with higher connectivity, lower latency, higher spectral efficiency, and relaxed channel feedback.
  • A. Fundamentals of NOMA: NOMA faces unresolved challenges including dynamic user pairing, transmission distortion, and channel and interference estimation before deployment in real networks.
  • B. Motivation to combine NOMA and MEC: MEC distributes computing resources from centralized clouds to the network edge, supporting massive connectivity and distributed computation near users.
  • B. Motivation to combine NOMA and MEC: Combining NOMA and MEC can improve user satisfaction and network performance while supporting low-latency transmission and complex 5G scenarios.
  • B. Motivation to combine NOMA and MEC: The survey identifies limited exploration of NOMA’s potential in MEC and reviews work combining the technologies for multi-user MEC systems.

C. State of the Art

Existing NOMA-MEC studies mainly optimize computation and communication resources, while the survey identifies broader open directions involving security, cooperation, and integration with mmWave massive MIMO.

  • C. State of the Art: Only a few studies examine MEC-NOMA scenarios, where multiple users can offload tasks simultaneously over the same frequency band.
  • C. State of the Art: Existing work studies partial offloading, power allocation, CPU frequency, task assignment, workload scheduling, and delay or energy minimization.
  • D. Learned Lessons and Potential Works: Open research directions include joint resource optimization, secure communications, cooperative NOMA-MEC, and coexistence with mmWave massive MIMO.
  • D. Learned Lessons and Potential Works: Joint optimization must coordinate offloaded workload, local computation, processing capacity, transmit power, latency, and energy consumption.
  • D. Learned Lessons and Potential Works: NOMA-MEC security is challenging because successive interference cancellation can expose another user’s message to decoding during simultaneous offloading.
  • D. Learned Lessons and Potential Works: Cooperative NOMA-MEC can use a helper MEC server as a relay, while mmWave massive MIMO can target higher data rates, connectivity, and computing capability.

IV. MEC WITH ENERGY HARVESTING AND WIRELESS POWER TRANSFER

Energy harvesting and wireless power transfer can complement MEC by supplying energy to constrained devices and expanding computation-offloading options. Their integration introduces challenges in energy variability, resource allocation, latency, and security.

  • Fundamentals: Energy harvesting captures ambient energy to power energy-constrained devices and prolong their lifetime.Harvestable sources may be natural or human-made, controllable or uncontrollable.
  • Fundamentals: RF-based wireless power transfer supports long-distance energy delivery and can combine wireless communication with power transfer.RF-based WPT uses far-field electromagnetic waves over distances of hundreds of meters.
  • Integration with 5G: Integrating EH/WPT with 5G systems can improve energy and spectral efficiency across IoT, D2D, HetNet, and cognitive-radio networks.The integration also requires approaches that address unstable ambient resources and the balance between harvested and consumed power.
  • EH/WPT-MEC Integration: EH/WPT can power MEC devices, enlarge offloading options, and support persistent operation of large populations of IoT sensors.MEC can also learn time-varying energy-source properties and reduce device processing time through workload offloading.
  • Challenges: Efficient EH/WPT-MEC operation remains constrained by communication and computation allocation, latency minimization, security, harvesting-process constraints, and server load imbalance.Renewable-powered servers may experience uneven energy availability and overload because harvested power varies across space and time.

C. State of the Art

Research on EH/WPT-MEC covers device-side and server-side energy architectures, offloading and scheduling, load balancing, and energy prediction. The surveyed literature identifies scalable resource management and prediction as continuing challenges, particularly for dense IoT networks.

  • Existing Schemes: Existing EH/WPT-MEC studies mainly examine energy harvesting or wireless power transfer at mobile devices and renewable-powered MEC servers.These schemes target WSNs, IoT, eRAC, and D2D systems, while server-side designs address costly or unavailable grid connections.
  • Existing Schemes: Representative studies optimize revenue, task schedules, offloading, energy consumption, execution delay, computation rate, age of information, and energy efficiency.Approaches include Lyapunov optimization, online dynamic scheduling, time allocation, beamforming, and joint communication-computation allocation.
  • Server-Side Optimization: Renewable-powered MEC servers require energy prediction and load balancing because unpredictable harvested power and limited computing capacity can overload individual servers.A distributed three-stage iterative algorithm alternates load balancing, channel allocation, and computation-resource allocation.
  • Open Challenges: Energy prediction increasingly uses machine learning and deep learning, while MEC can process collected data near the devices instead of sending it to a remote cloud.Implementation still involves data collection and substantial computation for high-dimensional big data.
  • Open Challenges: Dense IoT deployments require scalable offloading and resource-allocation designs across heterogeneous harvesting, WPT, and SWIPT devices and frequency bands.The designs must also consider impacts on the human body and environment.

V. MEC FOR UAV COMMUNICATIONS

UAVs provide flexible, mobile wireless infrastructure with coverage, line-of-sight, and emergency-response advantages. Combining UAVs with MEC creates aerial servers or enables UAV users to offload tasks, but requires joint mobility, communication, computation, and security optimization.

  • UAV Communications: UAVs support cost-effective deployment, line-of-sight links, coverage and capacity enhancement, and complementary connectivity during emergencies.Their mobility and maneuverability allow rapid reconfiguration and temporary service in situations such as disaster relief.
  • UAV-MEC Architectures: UAV-MEC systems have two main modes: UAVs can act as aerial base stations with MEC servers, or UAVs can offload tasks to ground MEC servers.The latter mode can use multiple ground base stations and more reliable line-of-sight links.
  • Challenges: UAV-MEC design must address three-dimensional deployment, flight-time and trajectory optimization, communication and computation allocation, and security.The required objectives can include relay minimization, energy-consumption minimization, and improved offloading opportunities.
  • Challenges: Jointly optimizing UAV mobility and resource allocation is needed because UAVs have limited flight time, size, weight, and power.Designs may additionally need to incorporate QoS, channel variation, delay, maximum speed, power allocation, and task assignment.

C. State of the Art

State-of-the-art UAV-MEC work studies aerial-server and cellular-connected-UAV configurations, including trajectory and task-scheduling optimization. The survey highlights performance analysis, energy-aware joint optimization, user association, and IoT integration as open directions.

  • State of the Art: UAV-MEC research considers aerial base stations with MEC servers and cellular-connected UAVs served by multiple ground MEC stations.In the cellular-connected configuration, UAV tasks are offloaded to selected ground stations during flight.
  • State of the Art: One study minimizes UAV mission completion time by jointly optimizing trajectory and computation scheduling under UAV speed and ground-server capacity constraints.The resulting problem is nonconvex, making a global polynomial-time optimum difficult to obtain.
  • Open Problems: UAV-MEC performance analysis should evaluate coverage probability, throughput, delay, and reliability against system design parameters.Three-dimensional deployment, short flight duration, and MEC delay requirements make this analysis challenging.
  • Open Problems: Energy-aware resource allocation remains open because UAVs have limited power and existing studies often optimize trajectory and resource allocation separately.Joint path-planning and resource-allocation design must also handle QoS, channel variation, delay, speed, power, and task constraints.
  • IoT Integration: IoT architectures use perception, network, and application layers, with MEC and fog servers potentially inserted for distributed computation or preprocessing.IoT deployments additionally face connectivity, interoperability, autonomic networking, security, privacy, and manageability requirements.

B. Motivation to use MEC for IoT and challenges

MEC supports IoT by processing data and providing computing capabilities near users, reducing traffic, latency, and device energy demands. Realizing these benefits requires addressing scalability, resource management, mobility, security, privacy, and trust.

  • Motivation: MEC gateways can aggregate and process IoT data at the network edge before it reaches the core network.This reduces infrastructure traffic and supports millisecond-range latency requirements for 5G Tactile Internet applications.
  • Challenges: Implementing MEC for IoT requires attention to scalability, communication, computation offloading, resource allocation, mobility management, security, privacy, and trust.
  • Application scenarios: MEC-enabled IoT research spans smart cities, smart homes, education safety, healthcare, and energy-management applications.Existing surveys and studies organize these works by application scenario and technical aspect.
  • Application scenarios: MEC-enabled IoT can provide local storage, real-time processing, and higher-level services for healthcare monitoring systems.

3) Vehicle-to-Everything (V2X) IoT:

MEC and 5G technologies support IoT scenarios including V2X, industrial systems, wearables, agriculture, and Tactile Internet applications. Reported studies associate edge processing with lower latency, resource use, energy consumption, and computation load.

  • Vehicle-to-Everything (V2X) IoT: 5G V2X use cases include information sharing, vehicle platooning, remote driving, cooperative collision avoidance, and dynamic ride sharing.Different V2X applications may require different data rates and communication ranges.
  • Vehicle-to-Everything (V2X) IoT: MEC-equipped cell towers and edge deep learning can support safer roads, smoother traffic flow, object recognition, and vehicle information sharing.
  • Industrial Internet: MEC supports latency-critical industrial IoT through TSN-based architectures, resource allocation, offloading, routing, and service-aware resource partitioning.Reported benefits include improvements in delay, successful response rate, fault tolerance, throughput, and energy efficiency.
  • AR, VR, and wearables: MEC and 5G are presented as addressing demanding AR, VR, and wearable applications through additional computing and storage resources, longer battery life, and low end-to-end latency.
  • Learned lessons: MEC-enabled IoT studies report 14% and 90% latency reductions, 12% lower radio resource consumption, 12.35% lower energy consumption, and 95% less transmitted data.
  • Challenges and future directions: Further work must address cooperation among dense MEC-IoT networks, service placement, limited edge resources, and integration of machine learning for optimization and privacy.

VII. MEC WITH HETEROGENEOUS CLOUD RADIO ACCESS NETWORK

H-CRAN combines heterogeneous radio access with centralized processing, while MEC adds computing near users. Their combination can improve coverage, energy efficiency, flexibility, deployment cost, and application support, but introduces substantial coordination challenges.

  • A. Fundamentals of Heterogeneous C-RANs: Network densification uses macrocells, small cells, and spatial spectrum reuse to address rising traffic and connected-device numbers.Dense HetNets improve coverage and capacity but face interference, energy-efficiency, flexibility, and scalability challenges.
  • B. Motivations and Challenges: H-CRAN provides coverage and energy efficiency, while MEC provides computing capability for low-latency applications.Collocating them can support more 5G applications using BBU-pool and RRH resources.
  • B. Motivations and Challenges: Collocating MEC with H-CRAN can reduce deployment investment by using existing C-RAN BBU pools or RRHs for additional task computation.
  • B. Motivations and Challenges: H-CRAN MEC offers operational flexibility, network re-configurability, broad coverage, energy savings, simplicity, and security through virtualization.
  • B. Motivations and Challenges: H-CRAN processing sites can be deployed across locations and can function as MEC servers for mobile-user tasks.Deployment design remains a challenge when processing power, site location, and application support must be balanced.
  • B. Motivations and Challenges: H-CRAN MEC requires joint management of wireless, backhaul, energy, computing, caching, and interference-related resources.Inter-carrier interference makes resource allocation more difficult than in traditional MEC systems.
  • B. Motivations and Challenges: Additional challenges include heterogeneous MEC-server deployment, user-to-BBU association, third-party application security, and integrity assurance.

C. State of the Art

State-of-the-art MEC research covers heterogeneous and C-RAN systems, machine-learning applications, and challenges in scalable resource, mobility, backhaul, and security management. The survey identifies distributed and adaptive approaches as important future directions.

  • C. State of the Art: Existing Het-MEC studies jointly optimize radio and computational resources under energy, transmit-power, latency, and computing-capability constraints.
  • C. State of the Art: C-RAN MEC research includes decentralized and centralized decision-making, resource allocation, transmit-power control, energy minimization, latency, and fronthaul constraints.
  • Open challenges: Centralized optimization can provide optimal or near-optimal solutions but is not scalable as mobile users, eNBs, and MEC servers increase.Lightweight and effective algorithms are therefore needed.
  • Open challenges: User mobility can require reassociation between RRHs and MEC servers, making dynamic user association, resource allocation, and service continuity important problems.
  • Open challenges: Dense spectrum reuse causes mutual interference that can reduce expected spectrum and energy efficiency in H-CRAN MEC systems.More sophisticated interference management is required to improve MEC-service QoS.
  • Open challenges: Limited backhaul or fronthaul capacity affects transmission time and users’ offloading decisions.
  • Open challenges: MEC applications sharing physical platforms with network functions create physical-security risks requiring security research and application integrity checks.
  • Machine learning in MEC: Machine learning is surveyed for edge caching, computation offloading, joint optimization, security, privacy, big-data analytics, and mobile crowdsensing.Its applications include learning from large data, online optimization, distributed algorithms, and federated learning with local data retention.

B. Machine Learning for Multi-Access Edge Computing

Machine learning is applied across MEC problems including caching, computation offloading, and resource optimization. Learning-based methods address dynamic environments and large decision spaces, but implementation can require substantial computational resources.

  • MEC optimization covers caching placement, radio and computing resource allocation, task assignment, and joint 4C optimization.
  • 1) Edge Caching: Edge caching research addresses where to cache, what to cache, and how to cache across network infrastructure and users.Caching locations include macro-eNBs, small-eNBs, and end users, while policies include LFU, LRU, preference-based, learning-based, non-cooperative, and cooperative approaches.
  • 1) Edge Caching: Deep reinforcement learning learns proactive caching policies for time-varying content popularity without predefined network models or explicit assumptions.Caching approaches either learn popularity estimation and policy separately or learn both simultaneously.
  • 1) Edge Caching: Learning-based caching consistently improves average cache hit ratio, training accuracy, and energy cost compared with baselines such as LFU and LRU.
  • Deep-learning implementation for MEC big-data analytics faces high computation demands, model-selection challenges, and parameter-optimization issues.
  • 2) Computation Offloading: Computation offloading is formulated using methods including Markov decision processes, DQN, hotbooting Q-learning, and deep learning.These methods optimize task distribution, offloading rates, bandwidth allocation, autoscaling, utility, payment, energy consumption, delay, and task loss probability.

3) Joint Optimization:

MEC research increasingly combines communication, caching, computing, security, big-data analytics, and machine learning. The survey identifies joint optimization and distributed learning as important directions while highlighting resource, privacy, and heterogeneity challenges.

  • 3) Joint Optimization: Joint 4C optimization targets communication, computing, caching, and control because conventional methods struggle with large action and state spaces.
  • 3) Joint Optimization: Research integrates networking, caching, and computing for fog-enabled IoT and connected vehicles, with deep reinforcement learning outperforming existing methods in one connected-vehicle framework.
  • Security and privacy are challenging because MEC combines diverse enabling technologies and distributed, heterogeneous edge environments.
  • Deep learning supports edge cyber-attack detection, while reinforcement learning supports anti-jamming offloading, physical authentication, and friendly jamming mechanisms.
  • 5) Big Data Analytics: MEC big-data analytics must address data volume, velocity, volatility, veracity, variety, distribution across finite servers, coupled 4C resources, and privacy.
  • 5) Big Data Analytics: Privacy-preserving big-data processing uses output perturbation and objective perturbation to protect training data through randomized query results or objectives.
  • 6) Mobile Crowdsensing: MEC-enabled mobile crowdsensing can parallelize and partition centralized large-scale problems, while hierarchical architectures divide learning and participant coordination across cloud and edge layers.
  • C. Challenges and Future Works: Machine learning is identified as useful for edge caching, computation, big-data analytics, security, and privacy, but distributed implementation remains an open problem.

IX. MISCELLANEOUS RESEARCHES

The survey reviews open-source activities, testbeds, and practical MEC implementations. It highlights ecosystem development, edge-platform fragmentation, and low-cost single-board computers as deployment options for edge services.

  • The ETSI WG DECODE seeks to accelerate MEC adoption and implementation through APIs, specifications, and ecosystem development.
  • Multiple edge-computing platforms can fragment the market, creating interoperability problems and limiting industry collaboration.
  • Single-board computers are considered efficient and cost-effective edge-cloud platforms because they offer low cost, low energy use, and sufficient resources for various applications.
  • Single-board-computer edge servers can support rescue missions where infrastructure has been destroyed by natural disasters.
  • Experiments with Raspberry Pi micro-clouds evaluate serving latency, hosting capability, memory read/write cost, and booting time.Reported evaluations found low latency and booting time while serving many users and reducing cost compared with an alternative.
  • MEC prototypes using OpenAirInterface and single-board computers support applications including streaming face detection, augmented-reality object analysis, object tracking, social sensing, and video analytics.
  • D2D-based MEC architectures use relay gateways as local clouds and direct connections to provide edge services to users and neighboring gateways.
  • Lightweight edge platforms combine single-board computers, lightweight virtual switching, and container virtualization while considering service QoS and hardware deployment cost.

3) Middleware for Edge Computing:

CloudAware illustrates how context-adaptive middleware can support computation offloading, while the survey situates middleware within broader MEC integration, implementation, and open challenges.

  • Middleware for Edge Computing: CloudAware predicts arbitrary context attributes to support computation offloading for applications operating over dynamic networks.It is identified as the first context-adaptive middleware for computation offloading.
  • Middleware for Edge Computing: 276% lower execution time was reported for CloudAware than local computing only, with the same offloading success rate.
  • Middleware for Edge Computing: MEC middleware research also addresses diverse QoS requirements, resilient data exchange, and application deployment across distributed heterogeneous edge systems.Approaches include user–broker connection orchestration, SDN-based network monitoring, and messaging middleware for emerging applications.
  • Middleware for Edge Computing: The survey covers MEC fundamentals, integration with forthcoming 5G technologies, lessons learned, open challenges, and future directions.
  • Middleware for Edge Computing: Integrating MEC with NOMA, energy transfer, machine learning, UAVs, and C-RAN can address connectivity, adaptability, scalability, sustainability, and service economics.
  • Middleware for Edge Computing: Open challenges include higher-level integration of currently independent technologies and coexistence of multiple MEC designs within a unified framework.
Loading 1906.08452v2…