PaperlayerView exact revision on arXiv

Source-linked AI summary

Survey on Multi-Access Edge Computing for Internet of Things Realization

Pawani Porambage, Jude Okwuibe, Madhusanka Liyanage, Mika Ylianttila, Tarik Taleb

arXiv:1805.06695v1cs.NI

TL;DR

IoT applications require infrastructure capable of handling growing data, computing, networking, mobility, and real-time demands. This survey synthesizes how MEC supports IoT through edge computing, gateway services, and enabling technologies, concluding that the two technologies are complementary for 5G and beyond while leaving several technical challenges open.

  • Problem

    IoT's anticipated growth creates demands that existing centralized cloud and network infrastructures may not adequately accommodate, especially for decentralized, mobile, latency-critical applications.

    Source: §I · p. 1, 1) Requirements · pp. 11, 14

  • Method

    The paper provides a holistic survey of MEC for IoT applications, technical requirements, enabling technologies, related projects, and future research directions.

    Source: §VII · p. 25

  • Results

    The survey identifies MEC as a gateway and edge-cloud technology for IoT, with benefits including reduced latency, lower traffic, computation offloading, and diversified network services.

    Source: §VII · p. 25

  • Takeaways & Limitations

    MEC and IoT are complementary technologies whose integration can support latency-critical IoT networks and contribute to 5G and future mobile networks.

    Source: §VII · pp. 25–26

Abstract

from arXiv · show

The Internet of Things (IoT) has recently advanced from an experimental technology to what will become the backbone of future customer value for both product and service sector businesses. This underscores the cardinal role of IoT on the journey towards the fifth generation (5G) of wireless communication systems. IoT technologies augmented with intelligent and big data analytics are expected to rapidly change the landscape of myriads of application domains ranging from health care to smart cities and industrial automations. The emergence of Multi-Access Edge Computing (MEC) technology aims at extending cloud computing capabilities to the edge of the radio access network, hence providing real-time, high-bandwidth, low-latency access to radio network resources. IoT is identified as a key use case of MEC, given MEC's ability to provide cloud platform and gateway services at the network edge. MEC will inspire the development of myriads of applications and services with demand for ultra low latency and high Quality of Service (QoS) due to its dense geographical distribution and wide support for mobility. MEC is therefore an important enabler of IoT applications and services which require real-time operations. In this survey, we provide a holistic overview on the exploitation of MEC technology for the realization of IoT applications and their synergies. We further discuss the technical aspects of enabling MEC in IoT and provide some insight into various other integration technologies therein.

I. INTRODUCTION

IoT is driving demand for highly capable, scalable network and computing infrastructures, while MEC extends cloud capabilities toward the network edge to support IoT services. The survey examines their complementary relationship, applications, enabling technologies, and technical challenges for 5G and beyond.

  • IoT is expected to generate substantial demand for data, computing resources, and networking infrastructure, requiring changes to existing networks and cloud technologies.Source: §I · p. 1
  • MEC extends cloud computing capabilities toward the edge of cellular networks, improving latency, bandwidth utilization, resource optimization, and access to network services.Source: §I · p. 1
  • Conventional centralized clouds face single-point-of-failure, location-awareness, reachability, and WAN-latency challenges for decentralized, mobile, geo-distributed IoT applications.Source: A. Role of MEC for IoT · p. 2
  • MEC and IoT mutually reinforce each other: MEC offloads computation from constrained IoT devices, while IoT expands MEC services across smart objects and vehicles.Source: A. Role of MEC for IoT · p. 2
  • MEC-IoT integration is motivated by lower infrastructure traffic, reduced application latency, and diversified network-service scaling, with latency emphasized as the most significant benefit.Source: A. Role of MEC for IoT · p. 2
  • The survey addresses MEC-IoT applications, scalability, communication, computation offloading, resource allocation, mobility management, security, privacy, trust, and integration technologies.Source: B. Paper motivation · p. 3

C. Paper organization

The paper organizes its survey around IoT application scenarios, MEC-enabled technical aspects, integration technologies, and future research directions. It also summarizes relevant surveys, acronyms, application characteristics, and MEC-IoT benefits.

Paper organization

Application scenarios

B. Healthcare

Healthcare, automotive, UAV, retail, wearable, and mixed-reality applications generate demanding IoT workloads involving latency, reliability, bandwidth, mobility, or computation. MEC addresses these demands through edge processing, local services, radio-resource control, and computation offloading.

Healthcare

  • Healthcare IoT includes wearable sensors and telemedicine, with some applications requiring extremely low latency for tactile feedback.Source: B. Healthcare · p. 6
  • MEC can process and store health data at edge components in hierarchical IoT healthcare architectures.Source: B. Healthcare · p. 6

Autonomous vehicles

Gaming, AR and VR

  • MEC supports AR and VR through low-latency offloading, increased tracking accuracy, expanded computational capacity, and on-demand cloud access.Source: D. Gaming, AR and VR · p. 7
  • Edge processing can locally prepare video frames, track positions, model environments, and identify objects for interactive VR applications.Source: D. Gaming, AR and VR · p. 7

Retail and wearables

  • Retail IoT applications include digital signage, supply chains, intelligent payments, smart vending, shelves, doors, streaming, and safety.Source: E. Retail · pp. 7–8
  • Wearable ecosystems face centralized-cloud latency limitations, while MEC combines edge and cloud infrastructures for delay-sensitive applications.Source: F. Wearable IoT (WIoT) · p. 8

G. IoT in Mechanized Agriculture

Agricultural IoT applies sensing, autonomous vehicles, remote monitoring, and analytics to production and management, while MEC assists with collecting and analyzing large datasets. Related industrial and smart-grid examples show similar edge-computing needs for automation, efficiency, and delay-sensitive operations.

IoT in Mechanized Agriculture

Related IoT domains

  • Smart grids integrate IoT devices to capture data for automation, but geographically distributed systems generate large data volumes requiring transfer, storage, and analysis.Source: H. Smart Energy · pp. 8–9
  • Traditional centralized cloud architectures are unsuitable for delay-sensitive smart-grid applications facing bandwidth bottlenecks and communication delays.Source: H. Smart Energy · p. 9
  • IIoT combines communication, automation, machine learning, and big-data analytics to improve manufacturing intelligence, connectivity, efficiency, scalability, and cost savings.Source: §I · pp. 9–10
  • MEC is positioned to support IIoT through real-time edge analytics, enhanced edge security, and mitigation of latency and resilience shortcomings.Source: §I · p. 10

III. TECHNICAL ASPECTS OF MEC ENABLED IOT

MEC-enabled IoT must address scalability, heterogeneous connectivity, communication-resource constraints, and interoperability across diverse deployment environments. The survey reviews these technical concerns and related integration technologies.

Scalability

  • Scalability depends on MEC-server compatibility with multiple network environments and the ability to support massive IoT deployments.Reviewed deployment settings include LTE macro base stations, 3G RNC sites, multi-RAT aggregation sites, and the core-network edge.Source: 1) Requirements · p. 10, 2) Related work · p. 10

Communication

  • MEC communication involves wireless offloading, backhaul access to remote clouds, and collaborative communication among IoT devices, edge hosts, and clouds.Wireless and backhaul links have limited capacity and can be unstable because of fading, interference, outages, and spectrum shortages.Source: 1) Requirements · pp. 10–11
  • Joint allocation of communication and computation resources is important because wireless and backhaul capacities must be shared alongside MEC-server computing resources.The survey identifies protocol redesign and cooperative allocation as challenges for integrating communication infrastructures with MEC and IoT.Source: 1) Requirements · p. 11

Interoperability

  • Heterogeneous LPWAN technologies create trade-offs among signal strength, operational range, throughput, and power consumption, making interoperability a major requirement.Examples include WCDMA, LTE, NB-IoT, Wi-Fi, Bluetooth, Zigbee, SIGFOX, and LoRA.Source: 1) Requirements · p. 11

Integration technologies

  • F-RAN consolidates heterogeneous networks into a 5G architecture, while C-RAN provides cooperative transmission but may introduce large latencies.The reviewed integration technologies also include mmW communications, OpenChirp, SDN, and NFV-based MEC implementations.Source: 2) Related work · p. 11

C. Computation Offloading and Resource Allocation

Computation offloading gives resource-constrained IoT devices access to additional edge computing power, but requires decisions that jointly account for delay, energy, bandwidth, connectivity, and server resources. The survey reviews full, partial, cooperative, and scalable offloading approaches.

Computation offloading

  • Computation offloading augments constrained IoT devices, potentially extending sensor battery life and reducing end-to-end latency for sophisticated applications.Offloading may be binary, fully executed at the edge, or partially split between the device and MEC server.Source: 1) Requirements · p. 12

Offloading decisions

  • Offloading decisions must consider execution delay, application partitioning, task dependencies, execution-time prediction, device capabilities, and connection quality.Realtime user input may need local processing, while other components can be offloaded according to radio, backhaul, and cloud conditions.Source: 1) Requirements · p. 12

Energy and latency

  • Energy-aware offloading balances execution and transmission energy against delay, especially for frequently transmitting or hard-to-reach IoT nodes.The trade-off applies to both full and partial offloading scenarios and is intended to extend battery life.Source: 1) Requirements · p. 12

Resource allocation

  • Joint resource allocation assigns single MEC servers to non-partitionable applications and multiple MEC servers to applications split into several parts.IoT devices may access MEC gateways through low-bandwidth low-power wireless connections.Source: 1) Requirements · pp. 12–13

Related work

  • An iterative bandwidth-allocation algorithm improved gateway-bandwidth utilization by more than 40% and battery life by up to 1.5 hours in a health-monitoring implementation.The approach addresses fragmentation caused by discrete, coarse-grained offloading levels at IoT end nodes.Source: 2) Related work · p. 13
  • Reviewed approaches also use nearby mobile devices, software load balancing, game theory, graph coloring, and graph-based algorithms to support scalable offloading.SDLB is reported to support about one million update requests per second.Source: 2) Related work · pp. 13–14

D. Mobility Management

Mobility management is crucial because movement away from an MEC node can degrade QoS or disconnect IoT devices. The survey reviews power control, VM migration, path selection, and energy-aware policies alongside security challenges.

Mobility management

  • Mobility management seeks to preserve QoS as mobile IoT devices move away from their computing nodes and experience increased latency or possible disconnection.The survey emphasizes ultra-reliable mobility management because many MEC-enabled IoT nodes are mobile.Source: 1) Requirements · p. 14
  • Energy-aware Mobility Management uses Lyapunov optimization and multi-armed bandit theories to optimize offloading delay under a long-term user-energy constraint.The scheme operates online without future system-state information and manages imperfect state information.Source: 2) Related Work · p. 14
  • The reported algorithms optimize delay while approximately satisfying the user's energy budget, but are ineffective in high-mobility scenarios typical of IoT networks.A connected node may move substantially during task processing, limiting the scheme's applicability.Source: 2) Related Work · p. 14
  • Power control can extend MEC coverage for UEs whose mobility is confined within a given space, while roaming beyond that region motivates VM migration or path selection.VM migration moves a virtual machine to a more effective computing node; path selection chooses a new communication path.Source: 2) Related Work · pp. 14–15, Figure 4 · p. 14
  • Without VM migration, the probability of connecting to the optimal MEC decreases as the number of hops between the eNB and UE increases, with additional delay also occurring.Migration is needed when an IoT node roams beyond the region extended by power control, where discontinuity and poor QoS risks increase.Source: 2) Related Work · p. 15

Security

  • MEC-enabled IoT deployments also face DoS, MitM, and VM-manipulation threats targeting networking, infrastructure, virtualization, and connected devices.DoS attacks can overwhelm critical resources, while MitM attacks can alter communications at the infrastructure layer.Source: 1) Requirements · p. 15

2) Related Work:

MEC addresses IoT privacy concerns by moving caching, processing, and analytics closer to data sources, while broader deployment must account for regulatory interoperability, data portability, security, and trust. The survey also identifies resource constraints and limited attention to trust management.

Privacy

  • MEC can process data near its source, reducing the burden on centralized clouds and core networks while limiting the raw data sent to central systems.The paper gives car-number-plate recognition as an example of edge processing that avoids transferring location information to centralized clouds.Source: 1) Issues and challenges · p. 16
  • Local MEC processing supports differentiated and jurisdiction-specific privacy policies and can reduce the impact of centralized data breaches.Only processed and selected data need to reach the centralized cloud for further processing.Source: 1) Issues and challenges · p. 16

Privacy governance

  • Privacy protection requires interoperable cross-border requirements, cooperation among jurisdictions, and data portability that supports new technologies such as MEC.The survey discusses global harmonization, compatibility with new technologies, and avoiding standards that prevent interoperability.Source: 1) Issues and challenges · p. 16

Trust

  • Trust is important for latency- and reliability-sensitive 5G IoT applications, yet it has received limited attention in recent research.Examples include remote surgery, emergency autonomous vehicles, factory automation, and tele-operated driving.Source: 1) Requirements · p. 16
  • Trust management must cover security, privacy, data sensing, data fusion, identity, secure communication, offloading services, and cooperation among edge servers.Trust relationships span IoT entities and enabling technologies such as MEC.Source: 1) Requirements · pp. 16–17
  • Tamper-resistant trust mechanisms are challenging for tiny IoT devices because their limited resources constrain integration.The survey also identifies human-computer trust interaction as requiring further attention at the application layer.Source: 1) Requirements · p. 17

2) Related work:

The survey reviews trust, security, and integration technologies for MEC-enabled IoT, emphasizing NFV and SDN as complementary virtualization and control approaches. It also surveys related architectures and applications.

  • IoT trust research covers trust evaluation, frameworks, data perception, identity, privacy, communication, computation, user, and application trust.Source: 2) Related work · p. 17
  • MEC-IoT integration relies on SDN, NFV, ICN, and network slicing as enabling technologies.Source: §IV · p. 17
  • NFV virtualizes core networking functions and reuses virtualization infrastructure and management within MEC architectures.Source: A. Network Function Virtualization · pp. 17–18
  • NFV can host VNFs and MEC applications together, while dynamically scaling network resources to improve MEC application scalability.Source: A. Network Function Virtualization · p. 18
  • Research integrates NFV and MEC for resource placement, low-latency multimedia, mobile gaming, HD video streaming, augmented reality, and local content caching.Source: A. Network Function Virtualization · p. 18
  • SDN separates network control into software-based controllers, and MEC can place those controllers closer to data-plane devices to reduce packet-processing latency.Source: B. Software Defined Networking · p. 18
  • SDN-MEC research targets flexible IoT services including reliable heart-attack detection, security, mobile-edge cloud deployment, and smart-home network densification.Source: B. Software Defined Networking · p. 18

C. Information Centric Networking

The survey presents ICN and network slicing as complementary technologies for MEC-enabled IoT. ICN supports content-centric delivery and caching, while slicing separates shared infrastructure into application-specific logical segments.

Information Centric Networking

Network Slicing

1) SESAME: Small cEllS coordinAtion for Multi-tenancy and Edge services (June 2015 - Dec. 2017):

The surveyed projects demonstrate MEC, NFV, virtualization, slicing, interoperability, and edge-cloud architectures for multi-tenancy and diverse 5G-IoT applications.

SESAME

ANASTACIA

  • ANASTACIA develops trust and security by design for heterogeneous, distributed, dynamically evolving cyber-physical systems using IoT and virtualized cloud architectures.Source: 2) ANASTACIA: Advanced Networked… · p. 21

5G-MiEdge

5G!Pagoda

  • 5G!Pagoda develops on-demand virtual mobile networks, scalable slicing, and programmable control for IoT and human-communication use cases.Source: 4) 5G!Pagoda · p. 21

INTER-IoT

5G-MoNArch

5G ESSENCE and MATILDA

9) 5GCITY (June 2017 - June 2019):

The surveyed projects and lessons learned connect edge virtualization, IoT pilots, and MEC capabilities to diverse 5G use cases. The survey identifies resource management and machine learning as continuing research needs.

5GCITY

  • 5GCity transforms infrastructure in Barcelona, Bristol, and Lucca into distributed, 5G-enabled edge-virtualization domains for municipalities, providers, and vertical sectors.Source: 9) 5GCITY (June 2017 - June 2019) · p. 21

MONICA

AUTOPILOT

5G-CORAL

  • 5G-CORAL uses virtualized networking and computing at the radio-access edge to blend functions, context-aware services, and applications for access convergence.Source: 12) 5G-CORAL: A 5G Convergent… · p. 22

Lessons learned

  • The survey identifies scalability, communication, offloading, resource allocation, mobility, security, privacy, trust, and standardization as MEC-IoT research areas.Source: §VI · p. 22
  • MEC supports IoT bandwidth and ultra-low-latency demands by preprocessing massive data at the edge, reducing bandwidth consumption and enabling faster responses.Source: 1) Lessons learned · p. 22

Future research directions

  • Efficient distribution and management of edge storage and computing resources remain necessary to realize MEC's full IoT benefits.Source: 1) Lessons learned · p. 22
  • Machine learning at the network edge may support adaptive decisions for data-intensive IoT applications, but autonomous-vehicle deployments still need further research and unifying theories.Source: 2) Future research directions · p. 22

B. Scalability

Future MEC-enabled IoT networks must address scalability, massive real-time data generation, heterogeneous access technologies, and the lack of standardized offloading mechanisms. Proposed directions include IPv6 adoption, SDN/NFV-based architectures, edge caching, and improved resource-management strategies.

Scalability

  • Real-time IoT sensing will generate huge volumes of readings, often with data lifespans of about 2 seconds, requiring refined scalability and data-management paradigms.Timeliness is critical because generated data may quickly lose value.Source: 1) Lessons learned · p. 22
  • IPv6 adoption is identified as a significant step toward advancing scalability in MEC-enabled IoT applications.Source: 2) Future research directions · p. 22
  • CONCERT combines NFV and SDN principles to enhance scalability, while MEC-server placement affects latency because core-network locations increase device distance.Source: 2) Future research directions · p. 23
  • Edge research addresses trade-offs between transmission rate and MEC-host storage, while heterogeneous IoT technologies create conflicting energy, throughput, and coverage goals.Source: 2) Future research directions · p. 23
  • Most offloading proposals rely on theoretical analysis or simulations, and no proper standard offloading mechanism yet exists for IoT-MEC systems.Source: 1) Lessons learned · p. 24

2) Future research directions:

Future MEC-enabled IoT research must improve mobility-aware offloading, migration, security, and privacy across heterogeneous and distributed infrastructures. The survey highlights prediction-based resource management, holistic mobility schemes, universal security standards, and updated privacy mechanisms.

Mobility and resource management

  • Mobility-aware resource management and computation offloading require precise investigation, alongside seamless offloading for scalable large-scale IoT deployments.Source: 2) Future research directions · p. 24
  • Prediction techniques can estimate offloading necessity, channel quality, and total cost under varying movement conditions.Source: 2) Future research directions · p. 24
  • VM migration may not suit highly delay-sensitive real-time applications, motivating holistic schemes combining power control, migration, compression, and path selection.Source: 2) Future research directions · p. 24

Security

Privacy

  • Existing privacy objectives are not compatible with current MEC, IoT, and 5G technologies, requiring updated privacy directives.Source: 1) Lessons learned · pp. 24–25

2) Future research directions:

The survey identifies unresolved integration challenges involving privacy, trust, standardization, heterogeneous ecosystems, and security regulation. It concludes that MEC and IoT complement each other while still requiring coordinated technical and institutional development for 5G evolution.

Privacy

  • Software Defined Privacy, Privacy by Design, and SDN-based privacy-aware routing are proposed to address privacy protection during or after MEC-IoT integration.Source: 2) Future research directions · p. 25

Trust management

  • Trust management remains under-investigated, with mutual trust among MEC servers needed for secure sharing of IoT datasets.Source: 1) Lessons learned · p. 25
  • A comprehensive context-aware trust framework is still lacking for holistic trust management across IoT perception layers and MEC edge servers.Source: 2) Future research directions · p. 25

Standardization

Regulation

Conclusions

  • MEC servers can act as gateways between latency-critical, massive IoT networks and the core network, providing edge-cloud computing and networking functions.Source: §VII · p. 25
  • MEC and IoT are described as complementary technologies with potential to advance 5G networks and beyond, alongside research needs in congestion control, latency-aware routing, and dynamic application routing.Source: §VII · p. 26

Related papers

Loading 1805.06695v1…