Raw data is worth nothing in itself: its value comes from a life cycle: collect, store, clean, transform, analyse, decide.
2.1What is data?
We distinguish structured data (tables), semi-structured data (JSON, logs) and unstructured data (text, images, video). The latter make up most of the volume and much of the raw material of generative AI.
2.2Databases: SQL and the others
- Relational (SQL): tables, relations, ACID transactions. The default choice when consistency matters: bank accounts, orders, records.
- NoSQL: documents, key-value, graphs. Chosen for volume, speed or flexibility, sometimes at the cost of less rigour.
- The key question: never "SQL or NoSQL?", but "what guarantees does my use require?"
2.3The modern data pipeline
- Collection from applications, sensors, the web, logs.
- Storage in a data warehouse, a data lake or a lakehouse.
- Transformation: cleaning, joining, modelling; preparation often takes up most of the time.
- Analysis: dashboards, statistics, AI models.
- Governance: knowing what data exists, who accesses it, where it comes from.
2.4The cloud: renting computing like electricity
The cloud is computing that is on demand, pooled, elastic and billed by use. Three levels: IaaS (machines and storage), PaaS (managed services), SaaS (finished software).
Under the hood: virtualisation, containers, regions, availability zones, automation. A crucial point inherited from cybersecurity: the shared responsibility model: the provider secures the cloud, the customer remains responsible for what they put in it.
2.5Who should run your IT?
- Public, private, hybrid, sovereign: each model trades off elasticity, control, cost and legal constraints.
- Repatriation: with a stable workload, some companies prefer to own their servers; hence FinOps, the discipline of optimising cloud spending.
- Edge computing: processing data close to its source for latency, cost or privacy.
Key takeaway
Data and the cloud rest on the same economic principle: pooling to turn fixed costs into variable costs. The counterpart is a technical, financial and legal dependence to be weighed case by case.
Sources
NIST SP 800-145, definition of cloud computing; Flexera, State of the Cloud Report; AWS/Azure/GCP documentation on the shared responsibility model.
Finished this module?
Your progress is saved on this device.