1. Why edge computing? #
Addressing latency challenges in an ultra-connected world.
The end of full centralization #
Sending all the data from a self-driving car to a central cloud takes too long (latency). Edge computing moves computing power as close as possible to the data source: on the sensor or in a local mini data center.
2. Edge vs cloud: complementary #
The edge handles real-time decisions (immediate reaction). The centralized cloud processes aggregated data for training AI models or historical archiving. The two work in tandem.
3. IoT, 5G and Industry 4.0 #
The arrival of 5G boosts edge computing by enabling massive throughput. In smart factories, the edge can detect an anomaly on an assembly line in a few milliseconds, avoiding costly accidents.
Hands-on: deploying as close as possible to users #
Edge and latency.
Measuring the value of the edge #
# Mesurer la latence depuis un client
curl -o /dev/null -s -w "DNS: %{time_namelookup}s connexion: %{time_connect}s TLS: %{time_appconnect}s premier octet: %{time_starttransfer}s\n" https://app.exemple.fr
mtr -rwzc 20 app.exemple.fr
# Déployer une fonction à la périphérie (Cloudflare Workers)
npm create cloudflare@latest edge-demo && cd edge-demo && npx wrangler deploy
# Edge sur Kubernetes léger
curl -sfL https://get.k3s.io | sh - # k3s : cluster Kubernetes de quelques centaines de Mo pour l'edgeEdge-to-cloud solutions and modular data centers #
Industrializing the edge.
Offerings #
| Approach | Examples | For |
|---|---|---|
| Cloud extended on premises | AWS Outposts, Azure Stack Edge and Azure Local, Google Distributed Cloud | The same APIs as the cloud, as close as possible to the data |
| Lightweight Kubernetes at the edge | K3s, MicroK8s, KubeEdge | Stores, factories, remote sites |
| Modular data center | Prefabricated containers or racks (power, cooling, security built in) | Deploy capacity in weeks on an industrial site or in an isolated area |
An edge-to-cloud architecture processes and filters locally (latency, network outages, volume) and sends aggregated data, learning and centralized monitoring to the cloud.
Good to know: Monitoring an edge fleet relies on local collectors that buffer during outages (OpenTelemetry Collector, Alloy, Bindplane) and on remote fleet management.
My progress
Sign in to track your progress and take notes. Sign in
💬 Questions about this lesson
No questions yet: be the first to ask for help or share your configuration.
Recommended Articles
Cloud deployment models: public, private, hybrid, community and multi-cloud
Understand cloud computing deployment models according to NIST: public, private, community and hy...
Serverless computing: definition, FaaS and use cases
Serverless computing explained: Function as a Service (FaaS), how it works, advantages, limits (c...
What is cloud computing? The NIST definition
Understand the fundamentals of cloud computing through the 5 essential characteristics defined by...
Stop moving to the Cloud, to micro-services...
I will certainly shock more than one person in this article, but don't hesitate to read to the en...
CaaS: state of the art and best practices
Container as a Service (CaaS) is a cloud computing model that lets companies manage and deploy ap...
IaaS, PaaS, SaaS: differences and examples of cloud service models
IaaS, PaaS, SaaS: what is the difference? Definitions, concrete examples (AWS, Azure, Google Clou...
Cloud best practices: security, resilience, costs and operations
Cloud computing best practices: pillars of a well-architected design, IAM and least privilege, mu...
Cloud native: definition and principles of a cloud native application
What is cloud native? The CNCF definition, the principles of a cloud native application (microser...
Controlling your AWS costs: pricing models, budgets and optimization
Understand and optimize the AWS bill: cost drivers (compute, storage, data transfer), Cost Explor...