Study Notes on Fog Computing at the Industrial level, Architecture, Latency, Energy, and Security

Conceptual Definition: Fog Computing (FC), a term originally defined by Cisco, refers to an extension of cloud computing that shifts resources from a centralized data center to the network edge. It focuses on enhancing frequent services, providing low latency, and enabling big data analysis at the point of origin.
Role in Industry 4.0: FC serves as a critical component of Smart Factories and Industry 4.0 (referred to as Advanced Manufacturing in North America). Its primary goals are to solve big data challenges, reduce energy consumption in industrial sensor networks, and improve real-time data processing, storage, and security.
The Challenge of Modern Infrastructure: Conventional cloud-centric systems struggle to meet the requirements of Industrial Internet of Things (IIoT). Cloud computing faces limitations regarding:
- High latency and bandwidth consumption.
- Traffic congestion and privacy concerns.
- Lack of location recognition and mobility support.
- Inability to manage the massive influx of heterogeneous devices.
Growth Statistics for Connected Devices:
- $2015$ : $15.41$ billion devices.
- $2017$ : $20.35$ billion devices.
- $2020$ (Expected): $30.73$ billion devices.

Cloud Computing: Considered a key enabler for IIoT, providing high-speed internet, storage, and computing power. However, it requires devices to send data and wait for responses, creating delays incompatible with real-time monitoring.
Fog Nodes (FNs): These are lightweight versions of cloud servers geographically distributed to provide resources closer to end devices. They are heterogeneous, consisting of:
- Servers and base stations.
- Access points and edge routers.
- Final devices (smart sensors and actuators).
The Internet of Things (IoT) vs. Industrial Internet of Things (IIoT):
- IoT: Focuses on managing large amounts of information delivered to users. IoT devices are often limited in resources and are relatively "vulnerable" to security threats.
- IIoT: Requires higher levels of reliability, latency control, and security. It utilizes closed-loop systems to collect information and update parameters in real-time. Benefits include failure prediction, resource planning, and quality optimization.
Industrial Revolution Progress: Transition from earlier stages to Industry 4.0. Currently, approximately one-fifth of industrial companies have digitalized processes; this is expected to rise to $85\%$ within five years.
Wireless Sensor Networks (WSN): Essential for gathering environmental data. In industrial contexts, these are called Industrial WSNs (IWSNs). They face strict energy resource limitations (battery life) and diverse topology challenges.

Layered Structure: FC architecture is typically divided into three distinct layers: the Cloud Layer, the Fog Layer (deployed within local facilities), and the Device Layer.
Scaling and Redundancy: Nodes can be scaled internally (adding software/hardware) or externally (adding more nodes). This distribution of tasks improves both scalability and system redundancy.
Specific Architectural Studies:
- Standard Compliance: Reference architectures for Industry 4.0 center on the $ISO/IEC/IEEE\,2010:2011$ standard.
- Distributed Computing: Use of the Morlet wavelet method built for the $C28x$ real-time digital signal processor (DSP) using Embedded C. This allows devices with limited resources to participate in edge computing. Matlab is used for signal noise reduction and analysis.
- Deployment Algorithms: Use of bio-inspired approaches, such as the "monkey algorithm" and "genetic algorithm," to minimize total installation costs in logistics centers.
- Heterogeneity Management: Implementation of a data-oriented mechanism based on $zeroMQ$ to solve problems related to ubiquitous access in industrial applications.
- Osmotic Computing: A new model supporting efficient IoT service implementation at the network edge, showing potential for memory savings in applications involving deep learning and Exhaustible Artificial Intelligence (XAI).

Requirement: Industrial environments cannot tolerate high transmission or processing delays. For instance, process variables must be monitored continuously without interruption.
Optimization Techniques:
- Hybrid Approaches: Combination of diffuse learning and fuzzy inference systems to make data loading decisions based on past experiences.
- Load Balancing: Unmanned Aerial Vehicle (UAV) architectures use load balancing based on the number of idle and occupied nodes to minimize task latency.
- Four-Layer Outlines: A structure comprising cloud, clustering, data programmer, and device layers has shown more than $15\%$ performance gain in specific work scenarios.
- Smart Grids: Hierarchical architectures for Smart Local Grids (SLG) focus on reducing delays between machines, nodes, and fog servers.

Security Vulnerabilities: Techniques used in cloud computing are not directly applicable to FC due to its geographic distribution, mobility, and heterogeneity. Decentralized systems are inherently more vulnerable to attacks.
Threat Mitigation:
- Threat Intelligence Platform (TIP): Necessary for defending organizations operating across cloud, fog, and edge environments.
- FC-IDS: A Fog Computing Intrusion Detection System designed to resist Distributed Denial of Service (DDoS). It utilizes a hypergraphic grouping model based on the A priori algorithm to manage fog node resources.
- Resource Constraints: Access control via cryptographic protocols and authentication is effective but consumes significant node resources, which many industrial nodes lack.
- Identity Management: Use of unique identities and keys on each node for safe storage and efficient recovery, ensuring time-sensitive data is processed at the edge while non-time-sensitive data goes to the cloud.

Battery Restrictions: Industrial nodes often have limited energy resources. Researchers use fog nodes to predict data measurements, thereby reducing the frequency of device performance cycles to save power.
Protocols and Standards:
- MQTT: Message Queue Telemetry Transport is considered a stable standard for real-time sensor communication due to its many-to-many characteristics.
- Network Time Protocol (NTP): Not ideal for IIoT as it consumes excessive resources and increases latency in dynamic topologies.
- LEACH Protocol: Low Energy Adaptive Clustering Hierarchy is used to decrease consumption in WSNs.
Data Compression: Implementation of lossy and lossless compression on edge devices. Lossy methods are generally more efficient for energy savings.

Research Summary (Table 1):
- Gaolei et al.: Service Popularity-Based Smart Resources Partitioning.
- Chekired et al.: Industrial IoT data scheduling for enabling smart factories.
- Fu and Liu: Secure data storage and searching strategies.
- Kumar et al.: Focus on improving response time levels in smart grids.
Discussion on Software Defined Networking (SDN): SDN is identified as a strategy to improve architecture through dynamic relocation and node management, though it has not yet been widely implemented at the industrial level.
Proposed Future Standard: The development of Fog Computing under the $IEC-61499$ standard is recommended. This would provide necessary portability, interoperability, and application reconfiguration specific to industrial norms.