LEARN. PRACTICE. GROW.Concept and mechanism
Structured data follows defined organization such as rows and columns; JSON documents can have flexible structure and optional properties. Audio and images remain unstructured even when their file catalog has well-defined fields. Format and engine are different decisions. CSV supports tabular interchange, JSON and XML represent hierarchies, Parquet organizes data by columns, and Avro by records. Selection should consider readers, types, and access patterns. OLTP maintains operational state through short operations; OLAP favors historical exploration and aggregation. In a transaction, atomicity requires included changes to commit or roll back as one unit.
Guided application
In a fictional closing, a job can finish without loading every movement. Define counts, control totals, and rejection routing before handover. The DBA oversees database security, availability, and recovery. The data engineer builds and operates ingestion and transformation. The analyst works with business on meaning, models, and visualization. One person can hold multiple roles, but responsibilities still need identifying. To relieve heavy reporting on production, assess an analytical destination with explicit acceptable lag. The decision does not end at copying data: identify who confirms completeness, who fixes failures, and who accepts a partial result.
Input of 120,000 and load of 118,400 require explaining 1,600 rejections even with a green job.
Common pitfalls
Format as engine; metadata as content; technical success as completeness; cloud as absence of owners.
Related topics: Relational modeling, integrity, and SQL · Relational services and compatibility · Non-relational storage and access
Define the data contract and owners accountable for meeting it.
Reference: Structured semi-structured and unstructured data · DP-900 skills measured 2026-07-21