Fog Computing is a distributed computing model that extends cloud-style processing, storage and networking capabilities closer to end devices and data sources, particularly within Internet of Things and geographically distributed systems. The term was introduced to describe an intermediate computational layer between centralised cloud infrastructure and endpoint devices, with the intention of processing some workloads near where data is generated rather than transmitting all raw information to a distant cloud.
Fog nodes can include gateways, routers, local servers, access points, industrial controllers or other intermediate computing systems. Processing can occur across multiple levels, from devices through local fog nodes and regional infrastructure to central cloud services. The architecture can reduce latency, conserve bandwidth and enable local decision-making while still using central cloud resources for large-scale storage, coordination and analytics. Fog Computing is therefore a distributed placement model rather than a specific hardware platform or software product, and its exact implementation varies according to the application and environment.
Fog Computing developed largely in response to the growing scale of connected devices and the limitations of sending all generated information to central cloud data centres. Internet of Things deployments can produce large quantities of telemetry, much of which may be redundant or time-sensitive. If every sensor reading is transmitted to a distant data centre, network bandwidth and latency can become constraints. Fog architectures place computation at intermediate points so data can be filtered, aggregated, transformed or analysed locally.
The resulting information can then be sent to central systems for long-term storage, broader analytics or coordination. The term "fog" conveys the idea of cloud-like capabilities distributed throughout the network rather than concentrated in a remote location. Cisco played a prominent role in popularising the concept, while academic and industry research subsequently explored fog architectures for IoT, smart cities, industrial systems and connected infrastructure.
A characteristic of Fog Computing is its hierarchical or layered distribution of resources. Endpoint devices may generate raw information, local fog nodes can perform immediate processing, regional systems can coordinate information across multiple local areas, and central cloud systems can perform resource-intensive or long-term tasks. This hierarchy is not mandatory in every implementation, but it illustrates how processing can be placed according to application requirements.
Local nodes can execute rules that need rapid responses, while central systems can combine information from many locations. The architecture can also support intermittent connectivity because some decisions can be made locally rather than waiting for a remote service. Workloads can be assigned based on latency, available compute resources, data volume, privacy requirements and network conditions. This makes Fog Computing particularly relevant to systems in which data is geographically distributed and immediate local processing is valuable.
Fog Computing has applications in industrial automation, transportation, smart buildings, environmental monitoring, healthcare-related sensing, energy systems and smart-city infrastructure. An industrial site can process sensor information locally to identify abnormal conditions without sending every raw measurement to a distant cloud. A traffic-management system can aggregate information from local sensors and cameras to coordinate signals or identify congestion. Smart buildings can process occupancy and environmental data locally to control heating, ventilation and other systems while forwarding selected information to central analytics.
Connected vehicles can use nearby infrastructure to process information relevant to local traffic conditions. The exact application depends on the required latency, reliability, privacy and computational capacity. Fog Computing is especially useful when local decisions have operational value and when network transmission of every raw data point would be inefficient or unreliable.
Data management in Fog Computing must account for distributed ownership and potentially inconsistent local state. Fog nodes may temporarily operate without access to central systems, creating periods during which local data cannot immediately be synchronised. Applications therefore need mechanisms for buffering, reconciliation and duplicate handling. Data aggregation can reduce network traffic by converting high-volume raw measurements into summaries or events. Local preprocessing can also remove irrelevant information before transmission.
Security policies should determine which data can remain locally stored and which information can be transferred to central infrastructure. Data integrity is important because local manipulation could affect physical processes in industrial or infrastructure systems. Backup and retention strategies must also account for the possibility that fog nodes have limited storage and may be physically less protected than central facilities. Distributed data management is consequently one of the main technical challenges of fog architectures.
Fog Computing differs from conventional cloud computing primarily through the location and distribution of resources. Conventional cloud models concentrate large-scale infrastructure in central or regional facilities, while fog models distribute processing across intermediate network locations closer to endpoints. It also differs from pure edge computing, although terminology overlaps considerably. Edge computing generally refers to computation close to data sources or users, while fog computing historically emphasises an intermediate distributed layer between endpoints and central cloud systems.
A local gateway performing simple filtering may be described as an edge device or fog node depending on the terminology used by the architecture. Fog computing can therefore encompass multiple intermediate layers rather than requiring every computation to occur directly beside the endpoint. The distinction is useful conceptually but is not universally standardised, so technical descriptions of the actual deployment are often more informative than relying solely on the label.
Security in Fog Computing is complicated by the number and diversity of participating nodes. A central cloud environment may have tightly controlled physical access, whereas fog nodes can be deployed in factories, roadside infrastructure, buildings or other distributed locations. Nodes require strong identity and authentication so that compromised devices cannot impersonate trusted infrastructure. Communications between devices, fog nodes and central services should be protected against interception and manipulation.
Software updates need secure distribution and integrity verification because manual maintenance across many locations is impractical. Segmentation can limit the consequences of a compromised node, while least-privilege permissions prevent local workloads from obtaining unnecessary access to central resources. Monitoring must also handle intermittent connectivity and potentially heterogeneous hardware. Physical tampering is another consideration where equipment is accessible to attackers. These factors mean that security must be designed as a distributed property rather than assuming that the central cloud perimeter protects every fog node.
Fog Computing can provide lower latency, reduced bandwidth consumption, local autonomy and more efficient handling of geographically distributed data, but these benefits are accompanied by operational complexity. Many nodes increase the number of components requiring deployment, monitoring, patching, authentication and replacement. Hardware heterogeneity can make software portability difficult, while limited local resources constrain which workloads can run at intermediate locations. Distributed coordination can introduce consistency and fault-management challenges.
The model is therefore particularly valuable when local processing provides a clear advantage over central processing, rather than simply because more computing locations are technically possible. Modern cloud and edge platforms increasingly incorporate capabilities that overlap with earlier fog concepts, including distributed orchestration, edge services and local cloud regions. As a result, Fog Computing remains an important architectural concept for understanding distributed IoT processing even though contemporary products and platforms may use different terminology for closely related capabilities.
