Source-linked AI summary
Fog Computing: A Taxonomy, Survey and Future Directions
Redowan Mahmud, Ramamohanarao Kotagiri, Rajkumar Buyya
TL;DR
Fog computing addresses the difficulty of supporting real-time, latency-sensitive applications for geo-distributed IoT devices through an intermediate layer near those devices. This chapter surveys its structural, service, and security challenges, develops and applies a taxonomy to existing work, and identifies research gaps and future directions, while noting unresolved sustainability, interoperability, and distributed-deployment issues.
Problem
Centralized Cloud datacentres face the storage and processing demands of billions of geo-distributed IoT devices, producing congestion, latency, and poor QoS.
Method
The chapter surveys Fog-computing developments, analyzes challenges, presents a challenge- and feature-based taxonomy, and maps existing works to it.
Results
The survey classifies existing Fog-computing research and uses the analysis to identify research gaps and propose future research directions.
Takeaways & Limitations
The taxonomy organizes Fog research across structural, service, network, and security concerns, supporting analysis of approaches and limitations in the field.
Takeaways & Limitations
The literature provides narrow discussion of sustainable and reliable Fog computing, and unresolved issues remain in interoperable architectures and distributed application deployment.
Abstract
from arXiv · showhide
In recent years, the number of Internet of Things (IoT) devices/sensors has increased to a great extent. To support the computational demand of real-time latency-sensitive applications of largely geo-distributed IoT devices/sensors, a new computing paradigm named "Fog computing" has been introduced. Generally, Fog computing resides closer to the IoT devices/sensors and extends the Cloud-based computing, storage and networking facilities. In this chapter, we comprehensively analyse the challenges in Fogs acting as an intermediate layer between IoT devices/ sensors and Cloud datacentres and review the current developments in this field. We present a taxonomy of Fog computing according to the identified challenges and its key features.We also map the existing works to the taxonomy in order to identify current research gaps in the area of Fog computing. Moreover, based on the observations, we propose future directions for research.
1 Introduction
Fog computing places distributed computing, storage, and networking capabilities between centralized Cloud datacentres and geo-distributed IoT devices. The chapter surveys its challenges, taxonomy, existing research, gaps, and future directions.
- Fog computing extends Cloud-based services closer to IoT devices and sensors through an intermediate distributed layer.It can use nearby networking components with computing, storage, and networking capabilities.
- Centralized Cloud datacentres struggle with billions of geo-distributed IoT devices, contributing to network congestion, high latency, and poor QoS.
- Fog environments use components such as routers, switches, set-top boxes, proxy servers, and base stations near IoT devices.
- The chapter analyses Fog computing differences, resource architecture, service quality, security issues, and recent literature.
- It presents a challenge-based taxonomy, maps existing works to it, identifies research gaps, and proposes future directions.
2 Related Computing Paradigms
Fog computing is related to Cloud, Edge, Mobile Cloud, and Mobile Edge computing but differs in how and where computational resources are deployed. Its broader use of edge and core infrastructure supports multi-tier deployment for IoT.
- Cloud computing provides utility services such as IaaS, PaaS, and SaaS, but centralized datacentres may be far from end devices.
- Edge computing processes data closer to its source using end devices, edge devices, and edge servers.
- Mobile Edge Computing combines edge servers with cellular base stations and supports two- or three-tier application deployment.
- Mobile Cloud Computing supports remote execution because smart mobile devices face energy, storage, and computational constraints.
- Fog computing can use both edge and core networking components, enabling multi-tier deployment and service-demand mitigation for many IoT devices.
3 Challenges in Fog computing
Fog computing faces structural, service, security, scalability, QoE, context-awareness, and mobility challenges. These arise from heterogeneous distributed nodes, constrained resources, complex provisioning requirements, and difficult real-time protection.
- Fog nodes are heterogeneous and distributed, making programming-platform development and task distribution across IoT, Fog, and Cloud infrastructures difficult.
- Structural challenges: Selecting suitable nodes, configuring resources, and deciding deployment locations are vital but challenging under varying operational requirements.
- Structural challenges: Distributed or virtualized Fog nodes require techniques and metrics for inter-nodal collaboration and efficient resource provisioning.
- Service challenges: Service Level Agreements and Service Level Objectives are difficult to specify because cost, energy, applications, data flow, and network status vary by scenario.
- Security challenges: Fog computing is vulnerable to security attacks, while authenticated access, privacy, and data-centric integrity are difficult to ensure without affecting QoS.
- Service challenges: Service scalability, end-user QoE, context-awareness, and mobility support are difficult to handle during real-time interactions.
4 Taxonomy
The proposed taxonomy classifies Fog-computing research according to key features and associated challenges. It organizes infrastructure, provisioning, objectives, network applicability, and security while acknowledging that heterogeneous studies prevent straightforward performance ranking.
- The taxonomy classifies existing Fog-computing works according to identified challenges and associated features.
- Fog Nodes Configuration: Fog Nodes Configuration covers heterogeneous computational nodes and techniques for managing collaboration among Fog nodes.
- Resource/Service Provisioning Metric: Resource/Service Provisioning Metric covers factors used to provision resources and services efficiently under different constraints.
- Service Level Objectives: Service Level Objectives capture the objectives attained when Fog computing operates between Cloud datacentres and end devices or sensors.
- The taxonomy includes applicable networking systems and security issues considered across different Fog-computing circumstances.
- The review maps existing works to the taxonomy, but diverse environments, topologies, applications, resources, and targets make relative performance comparisons difficult.
5 Fog Nodes Configuration
Fog computing uses diverse, distributed node types to place compute, storage, and networking closer to end devices. Each configuration offers benefits but also introduces coverage, centralization, interference, cost, privacy, or QoS challenges.
- The literature identifies five Fog node types: servers, networking devices, cloudlets, base stations, and vehicles.
- 5.1 Servers: Fog servers provide virtualized compute, storage, and networking capacity at geographically distributed locations, but can limit execution-environment pervasiveness.
- 5.2 Networking Devices: Networking devices such as gateways, switches, and IoT Hubs extend Fog coverage, although their physical diversity complicates resource and service provisioning.
- 5.3 Cloudlets: Cloudlets provide highly virtualized middle-layer capacity for many end devices, yet structural constraints can leave them centralized even at the edge.
- 5.4 Base Stations: Base stations can support Fog extensions of CRAN and VANET, while dense deployment faces networking interference and high deployment cost.
- 5.5 Vehicles: Vehicles can serve as distributed, scalable Fog nodes, but privacy, fault tolerance, and QoS assurance remain challenging.
6 Nodal Collaboration
Fog nodes collaborate through cluster, peer-to-peer, and master-slave techniques. These approaches enable shared or coordinated execution, but scalability, reliability, access control, bandwidth, and networking overhead remain constraints.
- The literature specifies three nodal collaboration techniques: cluster, peer to peer, and master-slave.
- 6.1 Cluster: Clusters may be formed by node homogeneity, location, computational load balancing, or functional-subsystem priorities.
- 6.1 Cluster: Cluster collaboration exploits several Fog nodes simultaneously, but static clusters lack runtime scalability and dynamic formation depends on load and node availability.
- 6.2 Peer to Peer: Peer-to-peer collaboration supports hierarchical or flat interactions, proximity-based organization, processed-output exchange, and virtual-instance sharing.
- 6.2 Peer to Peer: Peer-to-peer node augmentation is simple and reusable, but reliability and access-control issues are predominant.
- 6.3 Master-Slave: Master-slave collaboration centralizes control of slave-node functions, processing load, resource management, and data flow in a master node.
- 6.3 Master-Slave: Hybrid collaboration combines master-slave, cluster, and peer-to-peer interactions, but real-time decomposition requires high-bandwidth master-slave communication.
7 Resource/Service Provisioning Metrics
Fog resource and service provisioning considers time, data, cost, energy, and context. These metrics reflect execution and communication demands, application and user conditions, infrastructure expenses, and energy constraints.
- 7.1 Time: Time-based provisioning metrics include computation time, communication time, and service-delivery deadline.
- 7.1 Time: Computation time depends on resource configuration and existing load, and helps identify application activity relevant to resource and power management.
- 7.2 Data: Data-centric provisioning considers input data size, data-flow characteristics, heterogeneous data architecture, semantic rules, and integrity requirements.
- 7.3 Cost: Cost metrics include networking, deployment, execution, security, user willingness to pay, and migration costs.
- 7.3 Cost: Networking cost reflects bandwidth requirements and associated expenses, including data uploading, inter-nodal sharing, and bandwidth-related latency.
- 7.4 Energy: Energy-aware provisioning addresses device consumption, Fog-Cloud energy-latency tradeoffs, carbon emissions, and the residual battery lifetime of end devices.
- 7.5 Context: User and application context, including user characteristics, mobility, network status, task requirements, and current load, affects provisioning decisions.
8 Service Level Objectives
Fog research targets service-level objectives spanning latency, cost, networking, resources, and application management. Proposed architectures, policies, and platforms address these objectives across heterogeneous, decentralized, resource-constrained environments.
- Existing Fog research proposes architectures, programming platforms, mathematical models, and optimization techniques for management-oriented SLOs involving latency, power, cost, resources, data, and applications.
- 8.1 Latency Management: Latency management keeps service delivery below an accepted threshold representing maximum tolerable latency or an application QoS requirement.
- 8.1 Latency Management: Latency-oriented approaches initiate nodal collaboration, distribute tasks between clients and Fog nodes, or select the lowest-latency Fog node.
- 8.2 Cost Management: Cost management distinguishes CAPEX and OPEX, addressing distributed-node deployment, node placement and number, and provider cost diversity for VM hosting.
- 8.3 Network Management: Network management covers congestion control, SDN/NFV support, and seamless connectivity across physically diverse entities.
- 8.3 Network Management: Distributed IoT interactions can congest the core network, while local Fog processing can reduce the requests forwarded to Cloud datacentres.
- 8.4 Resource Management: Resource management includes estimation, workload allocation, coordination, utilization balancing, and QoE-aware scheduling across heterogeneous, resource-constrained nodes.
- 8.5 Application Management: Application management requires scalable programming platforms and supports computation offloading for resource-constrained end devices and mobile applications.
9 Applicable Network System
Fog computing has been applied beyond IoT across mobile, radio, optical, power-line, content-distribution, and vehicular networking systems. These applications place Fog capabilities near network edges to support communication, services, or resource augmentation.
- Mobile Network: Fog computing has been explored in mobile networks, especially for compatibility with 5G networking.5G offers higher speed, greater signal capacity, and lower service-delivery latency than existing cellular systems.
- Radio Access Network: Fog-based radio access networks have been investigated as a complement to Cloud-assisted radio access networks.The cited work examines the potential of Fog computing in radio access networks.
- LRPON: Fog computing has been integrated with Long-Reach Passive Optical Networks for optimized network design.LRPONs support latency-sensitive and bandwidth-intensive home, industry, and wireless backhaul services over large areas.
- Power-Line Communication: Fog-enabled power-line communication has been discussed for electric power distribution in Smart Grids.PLC transmits data and alternating current simultaneously through electrical wiring.
- Content Distribution Network: Fog nodes have been used as content servers in Content Distribution Networks, enabling end users to access distributed content with minimal delay.Fog nodes are positioned across the network edge, shortening access distance to content services.
- Vehicular Network: Vehicles at the network edge have been considered Fog nodes in vehicular networks to support communication, data exchange, and resource augmentation.Vehicular networks enable wireless communication among vehicles and provide computational and networking facilities.
10 Security Concern
Fog computing faces security concerns because it operates between end devices and Cloud datacentres. The reviewed literature addresses authentication, privacy, data encryption, and DoS attacks.
- Security Overview: Fog computing has high security vulnerability because it resides between end devices or sensors and Cloud datacentres.Reported security concerns include user authentication, privacy, secured data exchange, and DoS attacks.
- Authentication: Authentication mechanisms in Fog services aim to resist intrusion and prevent unwanted access to pay-as-you-go services.The literature also considers device, data-migration, and instance authentication.
- Privacy: Privacy assurance is important because Fog processes data closely associated with users’ situations and interests.Privacy challenges have also been identified in Fog-based vehicular computing.
- Data Security: Encrypting data at Fog nodes is required when processed data containing sensitive information is forwarded toward the Cloud.One proposed Fog-node architecture includes a data-encryption layer.
- DoS Attack: Resource-constrained Fog nodes are vulnerable to DoS attacks that issue many simultaneous irrelevant requests and make useful services unavailable.The resulting request load can keep Fog nodes busy and degrade performance substantially.
11 Gap Analysis and Future Directions
The chapter identifies research gaps spanning contextual management, sustainability, reliability, deployment, power, tenancy, pricing, simulation, and Fog-compatible standards. It proposes these areas as directions for improving Fog computing.
- Context-Awareness: Many contextual-information aspects remain unexplored despite prior work using context to estimate resources.Future research can investigate techniques for applying environmental, application, user, device, and network context to resource and service management.
- Sustainable and Reliable Fog Computing: Existing literature provides narrow discussion of sustainable and reliable Fog computing.Open issues include QoS, service reusability, energy-efficient resource management, consistency, availability, secured interactions, and fault tolerance.
- Application Deployment: Large-scale applications may require modular development and distributed deployment because individual Fog nodes are resource constrained.The chapter notes that existing distributed-application programming and deployment approaches still leave related issues unresolved.
- Power Management: Power management is necessary because deploying more active Fog nodes to handle simultaneous requests can substantially increase total system power consumption.Fog nodes must respond to many service requests from end devices and sensors.
- Multi-Tenancy: Multi-tenant Fog-resource support and QoS-aware task scheduling have not been investigated in detail.Available Fog-node resources can be virtualized and allocated to multiple users.
- Pricing and Billing: Fog pricing and billing policies require further development because vertically arranged resources create expenses for both users and providers.The chapter states that Fog pricing differs significantly from Cloud-oriented policies.
- Simulation: Real-world Fog testbeds are expensive and often not scalable, while only a small number of Fog simulators are available.The chapter identifies development of efficient Fog simulators as a future research direction.
- Standards and Interfaces: Cloud-based services in Fog require modified standards and programming languages, along with efficient networking protocols and user interfaces.These developments are intended to support seamless management of many Fog connections.
12 Summary and Conclusions
The chapter surveys recent Fog-computing developments, organizes challenges and properties into a taxonomy, maps existing research to it, and proposes future research directions.
- The chapter discusses structural, service, and security challenges in Fog computing and presents a taxonomy based on key challenges and properties.
- The taxonomy classifies and analyzes existing works according to their approaches to addressing Fog-computing challenges.
- The chapter proposes promising research directions based on its analysis of the Fog-computing literature.