- The paper presents a comprehensive survey on fog computing, detailing state-of-the-art architectures and algorithms.
- The paper systematically categorizes literature into application-agnostic and application-specific domains with critical evaluations.
- The paper identifies key challenges, including heterogeneity, QoS management, and scalability gaps, setting directions for future research.
A Comprehensive Survey on Fog Computing: State-of-the-art and Research Challenges
The paper "A Comprehensive Survey on Fog Computing: State-of-the-art and Research Challenges" by Mouradian et al., provides an extensive review of fog computing, an innovative paradigm aimed at addressing the limitations of traditional cloud computing. The primary intent of fog computing is to extend cloud capabilities to the edge of the network, thereby addressing the latency issues and the specific localization requirements posed by Service Level Agreements (SLAs) and new applications.
Introduction
The authors introduce fog computing as an evolution of cloud computing with complementary characteristics and emphasize its role in enhancing the efficiency of latency-sensitive applications such as disaster management, smart grids, and content delivery services. The paper stresses that fog computing is not a substitute but a complement to cloud computing, enabling processing at the edge while maintaining cloud interaction capabilities.
Existing Surveys and Tutorials on Fog Computing
The literature reviewed includes several tutorials and surveys that cover the basic concepts, challenges, and potential applications of fog computing. Notable works include those by Vaquero et al., Yi et al., and Yannuzzi et al., which offer a high-level overview and identify challenges but lack a comprehensive evaluative approach towards existing contributions in fog computing. The paper by Mouradian et al. differentiates itself by offering a critical evaluation of architecture and algorithms in fog systems based on well-defined criteria and presents a thorough review of existing literature.
Literature Classification
The paper organizes the existing literature into two main categories: architectures and algorithms for fog systems. Each category is further broken down:
- Architectural Aspects: Application-agnostic architectures which cover provisioning, resource management, communication, and federation aspects. Application-specific architectures focus primarily on areas such as healthcare, vehicular networks, and smart environments.
- Algorithmic Aspects: Divided into computing, content storage and distribution, energy consumption, and application-specific algorithms.
End-User Application Agnostic Architectures for Fog Systems
Mouradian et al. identify various programming and resource management architectures:
- Mobile-fog by Hong et al. allows writing programs targeting diverse nodes across the cloud, fog, and IoT strata.
- Distributed Data Flow (DDF) by Giang et al. uses a directed graph for application topology, facilitating application deployment across the cloud and fog.
- Layer-based architectures by Yangui et al. emphasize extending PaaS architectures and utilizing the REST paradigm for interactions.
Resource Management in Fog Systems
Key works include:
- Resource migration by Bittencourt et al., which focuses on migrating VMs across fog nodes to support user mobility.
- Resource allocation by Agarwal et al., proposing algorithms that distribute workload between fog and cloud strata.
- Task scheduling by Cardellini et al., utilizing extensions of the Storm framework for distributed QoS-aware scheduling.
Architectures for Specific Applications
The paper highlights significant application-specific architectures for healthcare:
- General healthcare by Stantchev et al. involving sensor-based infrastructures.
- COPD support by Fratu et al. and F2C by Masip-Bruin et al., which demonstrate how fog computing can improve healthcare monitoring and emergency response times.
- Other relevant applications include vehicular networks (e.g., Vehicular Fog Computing by Hou et al.) and smart environments (e.g., smart grid applications by Yan et al.).
Algorithmic Contributions
The paper also reviews algorithms addressing:
- Resource sharing (e.g., Abedin et al. and Oueis et al., which propose algorithms for efficient resource allocation among fog nodes).
- Task scheduling (e.g., Zeng et al. and Deng et al.), using strategies like genetic algorithms and convex optimization to schedule tasks across fog and cloud nodes.
- Load redistribution and offloading (e.g., Hassan et al. and Ye et al.), proposing mechanisms for offloading tasks from mobile devices to fog nodes.
Challenges and Research Directions
The authors identify several challenges and propose future research directions:
- Heterogeneity: Design of semantic ontologies for unified resource descriptions.
- QoS Management: Enhanced SLA management techniques for dynamic, multi-provider environments.
- Scalability: Mechanisms for resource scaling across all strata.
- Mobility: Comprehensive solutions supporting mobile end-users and fog nodes.
- Federation: Federated composition mechanisms ensuring seamless execution across domains.
- Interoperability: Standards for signaling, control, and data interfaces across heterogeneous domains.
Conclusion
The paper by Mouradian et al. offers an insightful assessment of the current state of fog computing and highlights pivotal research areas. It emphasizes the need for robust architectural frameworks and algorithms to leverage fog computing fully and identifies both the gaps and opportunities for future developments in this domain. The survey stands as a crucial contribution to the field, providing a structured analysis and setting the stage for subsequent advancements.