Source-linked AI summary
The Internet of Things, Fog and Cloud Continuum: Integration and Challenges
Luiz F. Bittencourt, Roger Immich, Rizos Sakellariou, Nelson L. S. da Fonseca, Edmundo R. M. Madeira, Marilia Curado, Leandro Villas, Luiz da Silva, Craig Lee, Omer Rana
TL;DR
The paper addresses how IoT-fog-cloud infrastructures can handle growing, heterogeneous IoT data and application requirements beyond what centralized cloud computing alone may provide. It reviews infrastructure, management, and applications, and identifies unresolved challenges including energy evaluation, privacy, and real-time urban-data processing. Its supported conclusion is that fog and cloud can extend and complement cloud environments for IoT, while effective infrastructure remains challenging.
Problem
Growing IoT data volumes and heterogeneous application requirements expose limits of centralized cloud computing for real-time, low-latency, mobile, and edge-oriented applications.
Method
The paper conducts a literature review of IoT-fog-cloud infrastructure, management, and applications, covering networking, orchestration, resources, energy, locality, trust, and application domains.
Results
The review identifies fog and cloud as complementary infrastructure components for IoT data processing and storage, while documenting challenges that remain for effective deployment.
Takeaways & Limitations
Fog can place processing and storage closer to IoT data, reducing transfers and response times while supporting contextual, privacy-sensitive, and heterogeneous applications.
Takeaways & Limitations
The paper notes that IoT-fog-cloud energy impacts require weighing added technology energy pressures against savings in application domains.
Abstract
from arXiv · showhide
The Internet of Things needs for computing power and storage are expected to remain on the rise in the next decade. Consequently, the amount of data generated by devices at the edge of the network will also grow. While cloud computing has been an established and effective way of acquiring computation and storage as a service to many applications, it may not be suitable to handle the myriad of data from IoT devices and fulfill largely heterogeneous application requirements. Fog computing has been developed to lie between IoT and the cloud, providing a hierarchy of computing power that can collect, aggregate, and process data from/to IoT devices. Combining fog and cloud may reduce data transfers and communication bottlenecks to the cloud and also contribute to reduced latencies, as fog computing resources exist closer to the edge. This paper examines this IoT-Fog-Cloud ecosystem and provides a literature review from different facets of it: how it can be organized, how management is being addressed, and how applications can benefit from it. Lastly, we present challenging issues yet to be addressed in IoT-Fog-Cloud infrastructures.
1. Introduction
IoT growth is driving unprecedented data, storage, and processing demands that centralized cloud computing may not fully satisfy for heterogeneous, latency-sensitive applications. The paper reviews how fog and cloud can be combined across infrastructure, management, applications, and unresolved challenges.
- Motivation: IoT expansion is expected to connect virtually all objects, increasing demands for data transfer, storage, and processing.Connected devices generate diverse information, from environmental measurements to human behavior.
- Motivation: Cloud computing offers flexible, low-investment access to storage and processing but may not meet real-time, low-latency, and mobile application requirements.Centralized data centers can be distant from clients, requiring multi-hop transfers that introduce delays and consume bandwidth.
- Motivation: Edge-oriented infrastructures can improve response time and reduce bandwidth use while accommodating heterogeneous IoT applications.The paper motivates combining localized edge execution with cloud capabilities.
- Scope: The paper reviews IoT-fog-cloud infrastructure, management, and application aspects, including processing, networking, services, resource allocation, energy, data locality, trust, and business models.It also identifies challenges spanning infrastructure, management, and applications.
2. IoT, Fog, and Cloud: Basic Definitions
IoT generates heterogeneous data and application demands, while cloud computing supplies centralized, virtualized, on-demand resources. Fog computing extends this model toward the edge through a hierarchy that offers varied service levels, lower latency, and reduced bandwidth use.
- IoT: IoT devices are heterogeneous in protocols, energy needs, computing capacity, and mobility, making management across the communication and processing stack challenging.The expanding device population also produces unprecedented data volumes.
- IoT: IoT applications transform large raw datasets into useful information and knowledge while imposing different requirements on the computing system.Application diversity adds another source of heterogeneity.
- Cloud computing: Cloud computing provides virtualized, shared infrastructure through isolated virtual machines and containers in centralized data centers.Its service models include IaaS, PaaS, and SaaS, with deployment models such as public, private, hybrid, and community clouds.
- Cloud computing: Cloud services use SLAs and commonly charge by time, transferred or stored data, or requests, supporting on-demand provisioning, elasticity, and faster time to market.These characteristics reduce upfront capital expenditure in exchange for operational expenditure.
- Fog computing: Fog computing places a hierarchy of fog nodes, cloudlets, or micro data centers between IoT edge devices and the cloud.Higher levels generally provide greater capacity, while lower levels are closer to the edge.
3. Literature review
The literature review organizes the IoT-fog-cloud ecosystem into infrastructure, management, and applications. It uses this structure to examine connectivity and protocols, operational concerns, and application benefits.
- Review structure: The paper reviews three facets of the IoT-fog-cloud hierarchy: infrastructure, management, and applications.Figure 1 provides an illustrative overview of the topics covered within the infrastructure.
- Infrastructure and management: Infrastructure coverage includes cloud and fog computing, networking connectivity and protocols, and fog support for 5G.Management coverage includes orchestration, resources, services, energy, device federation, and data locality.
- Applications: The applications facet considers urban computing, mobile applications, and Industrial IoT, emphasizing how they can benefit from fog computing.These application categories are presented as distinct areas of review.
3.1. Infrastructure
The infrastructure combines edge IoT devices, hierarchical fog resources, and more distant cloud data centers to match application needs. The review covers connectivity, protocols, data aggregation, locality, 5G support, and infrastructure challenges.
- Cloud and fog: The IoT-fog-cloud infrastructure is three-tiered, with IoT devices at the edge, fog resources across the network, and cloud data centers farther away.Application components can be placed at different levels according to latency, computing capacity, and data locality.
- Cloud and fog: Fog hierarchy levels trade proximity and latency against processing and storage capacity, with higher cloudlets generally supporting more devices but incurring longer delays.The first fog level can provide a nearby offloading option with lower latency but limited capacity.
- Cloud and fog: Fog nodes may be logical or physical entities that combine computing, storage, and networking, including routers, switches, access points, cameras, and servers.Processing near data sources can improve performance and security for critical applications.
- Networking: Connectivity technologies vary by scenario, including wireless links for sensors and mobile devices, wired links for factories, and Internet connections to the cloud.WLAN, Bluetooth, ZigBee, cellular, Ethernet-like, and multi-hop technologies are among the examples discussed.
- Networking: Different IoT devices require adaptable protocols because heterogeneous devices may need communication within milliseconds.Protocol abstraction layers help connect devices and consolidate collected data.
- Data collection and protocols: Data collection and aggregation can protect privacy, reduce redundant traffic, improve energy use, prevent bottlenecks, and enhance accuracy by removing outliers and misreadings.Fog nodes can provide local access control, encryption, isolation, and contextual integrity.
- Data collection and protocols: Data locality keeps processing near data sources, enabling contextual handling and more efficient decisions as raw data becomes meaningful information.Local storage and processing can also improve response time, reduce network traffic, and facilitate authentication, authorization, and privacy.
- 5G support: Cloud-fog processing can combine data from multiple IoT services and apply machine learning, while supporting geo-distributed real-time processing and runtime adaptability for 5G.F-RAN is cited as an architecture combining communications and computing operations.
3.2. IoT and Fog Management
IoT-Fog-Cloud management must allocate heterogeneous resources, manage locality and energy, orchestrate dynamic systems, and establish federation trust across constrained devices and distributed domains.
- Resource Allocation and Optimization: Resource allocation maps applications to available infrastructure resources while optimizing one or more potentially conflicting objectives.Schedulers use application requirements and infrastructure characteristics as inputs and produce application schedules.
- Resource Allocation and Optimization: IoT increases scheduling difficulty because application requirements and infrastructure characteristics are highly heterogeneous, while scheduling is NP-Complete in general.
- Resource Allocation and Optimization: Fog scheduling must distribute heterogeneous data, jobs, and services across the fog/cloud hierarchy while deciding which services run at the edge or cloud.
- Services and Orchestration: Fog and edge systems introduce serverless and migratable-service approaches, but vendor-specific implementations limit generalization while vendor-neutral models support shared hosting approaches.
- Energy Consumption: No holistic assessment yet weighs IoT-Fog-Cloud energy costs against energy savings in application domains.
- Data Management and Locality: Fog data locality can improve response time, reduce network traffic, and support security and privacy by keeping data and processing nodes close.
- Federation and Trust: Federation requires trust among administrative domains, with cryptographic methods commonly used to establish identity and trust.
- Federation and Trust: Applying federation to fog and IoT requires computing resources for cryptographic or consensus operations, which IoT devices may lack.
3.3. Applications
Urban computing integrates heterogeneous data sources to understand and address city problems, while mobile IoT applications create offloading, mobility, and resource-management challenges.
- Urban Computing: Urban computing acquires, integrates, and analyzes heterogeneous data from urban sensors, vehicles, and people to address city problems.
- Urban Data Sources: Urban studies draw on physical sensors, statistical data, existing city infrastructure, and location-based social networks.
- Urban Data Sources: Physical sensor data can be difficult and costly to obtain, while statistical data may be unavailable for target locations and available in diverse formats.
- Urban Data Sources: Location-based social networks provide spatio-temporal urban data through voluntary user participation and include systems such as Foursquare, Waze, Instagram, and Twitter.
- Mobility and Offloading: Mobile IoT devices often have limited computing capacity or power, motivating offloading to reduce energy consumption and response time.
- Mobility and Offloading: Fog supports offloading and can migrate or replicate data and computation along predicted mobility paths, but prediction errors and multi-tier management remain challenges.
- Mobility and Offloading: Resilient mobile applications can use path splitting and multipath routing, although mobility may connect devices to different fogs along their paths.
4. Future directions
Future IoT-Fog-Cloud deployments require integrated management across 5G network and computing domains, standardized interfaces, and mechanisms for heterogeneous ownership and technologies.
- 5G Integration: End-to-end 5G network slicing requires resource management across wireless, optical, packet, fog, and cloud domains.Current network virtualization has not yet achieved integrated orchestration across all these domains.
- Standards and Management: Middleware and APIs are needed to communicate device requirements, capabilities, network conditions, and feasible quality-of-service guarantees.These interfaces would support fine-grained allocation and automated service-level agreements.
- Standards and Management: Future management mechanisms must handle heterogeneous wireless technologies and multiple ownership models spanning edge devices, fog, and cloud resources.
4.2. Serverless Computing
Serverless and microservice management across the IoT-Fog-Cloud hierarchy must adapt service placement and networking to changing context, quality-of-service requirements, and resource constraints.
- Microservices Management: Automatic microservice adaptation must account for deployment location, context, and resource constraints at each hierarchy level.
- Microservices Management: Service ranking can support multicriteria decisions when reconfiguring services to satisfy quality-of-service requirements.
- Network-Aware Reconfiguration: Microservice deployment and reconfiguration must address heterogeneous networks and may require coordinated network reconfiguration.
- Resource Allocation: Resource-allocation optimization becomes harder as device, application, topology, mobility, and requirement variables increase and change over time.
- Resource Allocation: Data-stream volume and velocity must inform allocation decisions because stream-processing requirements depend on operations and collection frequency, not only input size.
4.4. Energy Consumption
Energy consumption is an open challenge across the IoT-Fog-Cloud ecosystem, requiring coordinated hardware, software, and data-management approaches. Service placement can improve data locality, but deciding what to place at the edge and for how long remains unresolved.
- Energy Consumption: IoT proliferation and rising data production increase pressure to reduce energy consumption across hardware and software.Approximate computing is identified as one potentially useful hardware-oriented approach.
- Energy Consumption: Economical data management should assess whether all data must be generated, transferred, stored, or processed continuously.The proposed assessments connect data importance and frequency of use with alternative management strategies.
- Energy Consumption: Heterogeneous communication technologies make orchestration responsible for handling distinct networks and addressing schemes.The ecosystem includes cellular, wireless, wired, and radio-frequency technologies.
- Energy Consumption: Smart service placement improves data locality by locating services near the data they process.Open questions include which services belong on edge nodes and how long they should remain there.
4.6. Applying Federation Concepts to Fog Computing and IoT
Federation concepts could support governance and policy agreements across fog and IoT environments, but deployment models, standardization, scalability, and edge-device constraints remain open challenges.
- Applying Federation Concepts to Fog Computing and IoT: Fog and IoT federation may be simplified through deployment and governance models based on out-of-band information.Small federations with known, fixed IoT device types may require simpler arrangements.
- Applying Federation Concepts to Fog Computing and IoT: A standardized Fog/IoT Federation Manager profile could enable wider federation deployment, but it must address scalability as managed devices increase.The paper frames scalability as a persistent concern for larger federated environments.
- Applying Federation Concepts to Fog Computing and IoT: Federation management depends on cryptographic identity and trust mechanisms, yet resource constraints may prevent their use directly on IoT devices.Cryptographic support may need to stop short of the most resource-constrained edge devices.
4.8. Orchestration in Fog for IoT
Fog orchestration for IoT still faces open issues involving privacy, security, performance, monitoring, placement, and cross-layer analytics. These challenges are intensified by distributed deployments, high device density, mobility, and critical application requirements.
- Orchestration in Fog for IoT: Fog orchestration remains an open research area despite recent developments.The paper identifies multiple unresolved issues requiring further attention.
- Orchestration in Fog for IoT: Privacy mechanisms must address the GDPR and similar regulations because fog nodes near users gather, store, and process potentially sensitive data.The regulatory requirement applies to distributed fog deployments handling end-user information.
- Orchestration in Fog for IoT: Security mechanisms must prevent software, hardware, and network attacks against fog orchestrator nodes in distributed, dynamic, large-scale environments.The security challenge follows from the multiple perspectives created by the fog-based IoT setting.
- Orchestration in Fog for IoT: 5G fog orchestration must monitor dense, mobile systems while meeting critical applications' latency and reliability requirements.Component selection and placement directly affect performance in dynamic orchestration.
- Orchestration in Fog for IoT: Research should develop multi-level real-time analytics and efficient optimization mechanisms that operate across the fog orchestrators' layered structure.The paper calls for cross-layer solutions for multidimensional fog-based IoT data.
4.9. Business and Service Models
Fog computing introduces business and service-model questions beyond established cloud charging and billing practices. Its multi-stakeholder, hierarchical, and mobile setting makes resource allocation and service placement difficult, with a general model still unresolved.
- Business and Service Models: It remains unclear whether fog computing can adopt cloud business and service models or requires new ones.The paper presents this as an open question rather than a settled conclusion.
- Business and Service Models: Fog infrastructure management involves stakeholders ranging from autonomous academic and industrial systems to telecom operators and public authorities.Hybrid-cloud deployment can extend local resources with cloud resources.
- Business and Service Models: Resource allocation for mobile users must account for mobility patterns, diverse application requirements, and periods of cloudlet overload.The scenario is highly dynamic when many users concentrate in one location.
- Business and Service Models: Fog resource management must jointly decide data and computation placement while considering user speed and the hierarchy's migration costs.Higher-speed users could be assigned to higher-level cloudlets to reduce migrations.
- Business and Service Models: A resource-allocation model integrating user mobility, fog/cloud hierarchy, and application requirements remains an unresolved challenge.The paper identifies this integration as a central boundary for current approaches.
4.11. Urban Computing
Urban computing in the IoT-Fog-Cloud continuum faces open challenges involving large-scale, dynamic, and privacy-sensitive data, while industrial applications require interoperable frameworks for conflicting requirements.
- Urban Computing: LBSN-based urban computing must address temporal dynamics because static representations can lose relevant information.
- Urban Computing: Large LBSN datasets create real-time processing, storage, and indexing challenges beyond conventional data-processing and database systems.
- Urban Computing: LBSN exploration can threaten user privacy by enabling inference of preferences and behavior, requiring privacy protections.
- Industrial Internet of Things: Industrial Internet of Things software requires interoperable frameworks that accommodate varying and potentially conflicting user and system requirements.
- Industrial Internet of Things: The paper anticipates that no single framework may support every IoT-Fog-Cloud scenario.
5. Conclusion
The conclusion presents fog-cloud integration as a way to support IoT applications with diverse requirements and summarizes the paper’s discussion of infrastructure challenges. It also anticipates continuing resource-management challenges as device and application heterogeneity grows.
- Conclusion: Fog-cloud integration is presented as promising for supporting IoT requirements ranging from low-latency, real-time applications to storage- and processing-demanding workloads.
- Conclusion: The paper defines and discusses edge and fog computing scenarios and explains how they extend and complement established cloud environments for IoT applications.
- Conclusion: Resource management and efficiency are expected to remain challenging as the number of devices and heterogeneous applications continues to grow.