Fog Computing
Definition
Distributed computing architecture that extends cloud computing to the edge of the network, processing data locally before sending to the cloud.
Use Cases
- Cisco: Industrial IoT and smart infrastructure scenarios where data must be processed close to where it is generated (e.g., filtering sensor streams and making low-latency decisions). — Cisco popularized the fog computing concept and has delivered fog/edge-capable networking and compute platforms that place compute and analytics functions on or near network devices (gateways/routers) so data can be processed locally and only relevant results are forwarded to centralized systems. (Reduced backhaul bandwidth and improved response times for time-sensitive decisions by processing and filtering data nearer to devices instead of sending all raw data to a central cloud.)
- Siemens: Factory-floor monitoring and control where equipment telemetry needs near-real-time processing for quality checks and anomaly detection. — Deployed on-premises/edge industrial computing and gateway systems to aggregate and preprocess machine data locally, then forwarded summarized or selected datasets to central IT systems or cloud services for longer-term analytics and reporting. (Faster local detection of issues and reduced volume of data sent upstream, supporting more reliable operations in environments with strict latency and connectivity constraints.)
- NVIDIA: Smart city and video analytics use cases (traffic monitoring, safety analytics) requiring low-latency inference close to cameras. — Used edge AI systems to run inference near video sources, producing events/metadata locally and sending only relevant clips, alerts, or aggregated insights to centralized storage/analytics. (Lower end-to-end latency for real-time responses and lower network/storage costs by avoiding continuous upload of raw high-bitrate video streams.)
Frequently Asked Questions
- What's the difference between fog computing and edge computing?
- Edge computing usually means running compute directly on or very near the device (like a camera, sensor gateway, or on-prem server). Fog computing is a broader architecture that adds one or more intermediate layers between devices and the cloud (for example, gateways, local micro–data centers, or network nodes) to aggregate, filter, and process data before sending selected results to the cloud.
- When should I use fog computing?
- Use fog computing when you need low latency decisions, want to reduce bandwidth by filtering/aggregating data locally, must keep some data on-site for privacy or compliance, or operate in locations with unreliable connectivity. Common fits include smart cities, industrial IoT, retail analytics, connected vehicles, and healthcare devices where local processing is required even if the cloud link is slow or intermittent.
- How much does fog computing cost?
- Costs depend on the number of sites and the hardware/software you deploy near the edge. Key factors include: edge/fog hardware (gateways, rugged servers, accelerators), software licensing (device management, orchestration, security), connectivity (cellular/MPLS/ISP), operations (remote monitoring, patching, field support), and any cloud services used for centralized storage/analytics. Fog can lower cloud egress and storage costs by sending less raw data, but it typically increases distributed infrastructure and operational costs.
Category: emerging
Difficulty: advanced
Related Terms
See Also