Fog Computing: Enabling the Future of Connected Devices

This article is based on a piece originally published on cio.com.

What Is Fog Computing?

There is an IT architectural concept that combines roots stretching back to the 1800s with some of the most pressing infrastructure challenges of the modern era—and most engineers have never heard of it. It is called Fog Computing, and it has attracted founding consortium members including Dell Technologies, Cisco, Intel, and Microsoft.

In the simplest terms: Fog Computing is an IT systems architectural model that distributes computing, storage, and networking throughout the device-to-cloud landscape based on economic and physical constraints. Rather than routing all data from connected devices to centralised cloud endpoints, fog computing inserts a hierarchy of compute resources between the device and the cloud—closer to where data is actually generated.

Why Cloud-Only IoT Architectures Hit a Wall

Researchers at Cisco are widely credited with identifying the core problem: there is not enough public network capacity for tens of billions of devices to continuously stream data to a handful of large cloud providers. Physics and economics both get in the way.

Consider what a cloud-only IoT model requires at scale:

• Bandwidth: Projections for connected devices now exceed 50 billion. Sending raw sensor data from all of them to centralised cloud endpoints would require network capacity that simply does not exist today—and public networks are already under growing pressure from streaming media and entertainment.
• Latency: Many IoT applications—industrial monitoring, autonomous systems, healthcare devices—require near-real-time analysis. A round-trip to a distant cloud data centre introduces latency that can make responses too slow to be useful or safe.
• Reliability: Cloud-only models create a single dependency chain. If network connectivity is interrupted, edge devices lose their ability to act on data.
• Security: Transmitting sensitive sensor data across public networks to remote endpoints increases the attack surface significantly.

The conclusion is straightforward: IoT applications will need to compete for public network capacity on economic terms, and the economics increasingly favour computing resources located closer to where devices are installed.

A Concept With Deep Historical Roots

Remote monitoring and control of machines and environmental factors is not a new idea. Wikipedia’s article on Telemetry references an 1847 example in which French engineers built a network of weather and snow-depth sensors on Mont Blanc that transmitted real-time information to Paris.

Translated into modern fog terminology, that 19th-century system maps almost directly onto contemporary IoT architecture:

• The weather and snow-depth sensors → IoT endpoint devices
• A ruggedised local computer on the mountain → an edge gateway
• A private network connection → a low-latency, reliable link to a fog node
• The data centre in Paris → a fog node or cloud back-end for long-term analysis

The same pattern emerges across many related disciplines: machine-to-machine (M2M) communication, mesh computing, mobile edge computing, and dew computing. Each has its own standards bodies, communities, and best practices. Fog computing aims to serve as a unifying architectural umbrella for all of them.

How Fog Computing Works in Practice

Fog computing introduces a hierarchy of compute, storage, and networking resources between cloud data centres and end devices. This hierarchy typically looks like this:

1. Endpoint devices – Sensors, actuators, meters, cameras, and other IoT hardware that generate data.
2. Edge gateways – Local compute nodes (often ruggedised hardware) that aggregate data from nearby devices, apply filtering, run lightweight analytics, and pass only relevant data upstream.
3. Fog nodes – Mid-tier servers or clusters, potentially located at the network edge or in regional data centres, that perform more complex analytics, store intermediate data, and coordinate with both gateways and the cloud.
4. Cloud back-end – Centralised infrastructure reserved for long-running analytics, machine learning model training, archival storage, and management of the wider system.

By processing data at the appropriate tier, fog architectures reduce the volume of data that must traverse public networks, cut round-trip latency for time-sensitive decisions, and improve resilience when connectivity is intermittent.

Why Fog Computing Matters for DevOps and Platform Engineering

For DevOps and platform engineers, fog computing is not just an academic concept—it has direct operational implications:

• Pipeline design: Data pipelines must be designed to route data to the right processing tier, not just to a central cloud endpoint.
• Deployment targets: CI/CD pipelines may need to deliver software updates to edge gateways and fog nodes, not just cloud services.
• Observability: Monitoring and logging must span all tiers. An alert generated at an edge gateway that never reaches the cloud is invisible to centralised monitoring tools.
• Security posture: Each tier must be hardened independently. Edge devices running in uncontrolled physical environments require a different security model than cloud-hosted services.
• Interoperability: With many existing standards bodies and protocols in play—from industrial automation to consumer IoT—platform engineers must manage integration complexity across heterogeneous ecosystems.

The Road Ahead

The growth of IoT is real, rapid, and—under cloud-only architectural assumptions—unsustainable. Fog computing offers a principled, layered approach to distributing computation where it is physically and economically rational to do so.

The current landscape of related but fragmented disciplines (M2M, mesh computing, mobile edge computing, and others) each contribute valuable thinking, but lack the critical mass to rapidly scale IoT deployments to the level envisioned. A unified fog framework, implemented with open and interoperable standards, is a necessary foundation for any infrastructure team serious about building systems that can handle tens of billions of connected devices.

If you are designing IoT infrastructure today, the question is not whether fog computing principles apply to your architecture—it is how deliberately you are applying them.