Fog Computing and IoT

Fog Processors

• Sensors generate massive amounts of data in IoT, which presents data challenges.
• Sending all sensor data to the cloud requires high network bandwidth and may face capacity limitations.
• Cloud services may struggle with processing large data volumes from numerous IoT applications.
• Fog computing, introduced by Cisco in 2012, brings data processing and storage closer to sensors.
• The core idea reduces the need to send all data to the cloud.
• Key benefits include low latency, geographic distribution, lower operating expenses, and scalability.
• Low latency supports real-time services by processing data locally.
• Geographical and large-scale distribution offers distributed computing and storage resources.
• Lower operating expenses saves network bandwidth via local data processing.
• Scalability is achieved by bringing processing and storage closer to IoT devices.
• Fog computing acts as an intermediary layer between the cloud and end devices.
• Fog nodes are devices with computing, storage, and network connectivity.
• Examples of fog nodes are industrial controllers, switches, routers, embedded servers, and video surveillance cameras.
• Primary characteristics of fog computing include location awareness, and geographical distribution.
• Fog computing provides support for mobility, real-time interactions, and interoperability.
• Location awareness enables support for location-based services.
• Fog services and applications are distributed across multiple locations, unlike centralized cloud services.
• Direct connectivity to mobile devices facilitates mobility.
• Real-time interactions occur between fog nodes, unlike the batch processing in cloud computing.
• Fog nodes from different manufacturers can work together across different domains and service providers, meaning they're interoperable.

Hierarchical Fog Infrastructure

• The levels of the infrastrucutre are monitoring and control, operational support, and business support
• Monitoring and Control (Tier 1) is closest to IoT devices and reacts to sensor-generated data, adjusting device states.
• Tier 1 offers very low storage, covers a very local area, and provides the fastest response, supporting real-time actions.
• It senses problems and commands actuators, supporting location-based services with geographical location sharing.
• Operational Support (Tier 2) consists of aggregation nodes that aggregate, analyze, filter, compress, and transform data from Tier 1.
• Tier 2 covers a larger area, has higher latency compared to Tier 1, and includes smart intermediate devices with storage, computation, and routing.
• Business Support (Tier 3) is closest to cloud services and turns aggregated data into actionable insights for enterprise-level services.
• Tier 3 has the highest storage and analytical capabilities.
• It's a centralized tier with high latency, handling data transmission from all lower tiers and composed of high-end servers and data centers.
• Data is transmitted to Tier 3 only when necessary, enhancing efficiency in resource allocation and packet forwarding.
• Fog computing hierarchy can be scaled based on business needs for storage, network connectivity, and analytics.
• The architecture is not mandatory; the functions of each tier can be performed by different or consolidated devices.
• Some use cases may implement all three tiers in the cloud, while others might only use fog for monitoring and control, with the remaining functions handled by the cloud.

Deployment Scenarios

• Entirely Cloud-Independent Scenario: All three tiers are deployed on fog nodes, suitable for extremely latency-sensitive applications.
  • Examples: healthcare, ATM banking, and armed forces combat systems.
  • Critical for applications involving human lives and when cloud availability may be restricted.
• Hybrid Scenario (Fog in First and Second Tiers): Uses fog nodes for the monitoring and control (first tier) and operational support (second tier) layers.
  • Suitable for latency-sensitive applications, exemplified by commercial building management, solar panel monitoring, and retail.
• Hybrid Scenario (Fog in First Tier Only): Uses fog nodes only for the monitoring and control (first tier) layer.
  • Suitable for less latency-sensitive applications, such as commercial UPS device monitoring, mobile network acceleration, and content delivery networks (CDNs) for Internet acceleration.
• Fog-Independent Scenario: No fog nodes are used; the entire architecture relies on cloud services.
  • Feasible for use cases where using fog might not be economical or practical, such as agriculture, connected cars, and remote weather stations.
• Businesses can add more layers to the fog computing hierarchy for extra intelligence and deeper data analysis.
• Fog nodes have built-in intelligence to redirect data to other fog nodes or to the cloud, depending on current processing load and tier capabilities.

Fog Computing Implementation

• Terminal Nodes: Deployed on IoT sensors, sensing and sharing geographical location.
• First tier: Virtual Clusters (VCs): Localized terminal nodes form VCs, seen as a unit by the higher tier.
  • Each VC gets connected to its geographically nearest fog instance (FI).
  • Data transmission is possible between both tiers with edge gateways.
• Second tier: fog instance (FI): hierarchy is composed of fog-aggregation nodes.
  • Can be deployed on networking devices that bear some computing, analytical, and storage feature, e.g., routers and switches.
  • FIs are able to transfer data to the next and final tier, the cloud tier, through gateways.
• Tier 3: Cloud Tier: Consists of cloud gateways connected to data centers (DCs) that handle large-scale data processing and storage.
• Software-Defined Networking (SDN) can be used alternatively.
• SDN is designed to ensure system flexibility
• Adapts a control plane within the network that separates control functions from data-forwarding functions
• The separation makes packet forwarding more efficient, flexible, and scalable while providing better quality of service (QoS) management for enterprises.
• Data plane: switches forward data packets
• Control plane: carriers network controller and management operator to control how devices in the data plane operate
• Through an application programming interface (API) of controller, the control plane is programmed to function in desired way affecting the whole network
• SDN-based Architecture
• Exploits massive number of cellular base stations (BSs) deployed in the mobile network on a global level
• Each BS could be directly connected to a fog node that aggregates data from local devices, processes some of it, and sends rest through SDN-Cellular core to the cloud
• Each BS group of a particular location could be connected to an edge device in the network core
• Edge device could be also connected to a fog node to aggregate data from the localized base stations that are connected to it
• SDN-based architecture adopts OpenFlow switches in the network and an OpenFlow controller manages the forwarding plane of the base stations and switches while monitoring the network traffic via the OpenFLow Protocol (developed by the open networking foundation (ONF)).
• Controller can be reprogrammed through an API by businesses, which provides flexibility of functioning