Data Warehouse vs Data Lake

Data warehouse и data lake — разные подходы с разными целями. Понимание trade-offs критично для правильного выбора.

Data Warehouse

Structured data, defined schema (schema-on-write). Optimized для SQL analytics. Columnar storage (Redshift, Snowflake, BigQuery). Часто expensive per TB, но fast queries. ETL pipelines: transform до загрузки. Ideal для business intelligence, reporting.

Data Lake

Все типы данных: structured, semi-structured, unstructured. Schema-on-read. Cheap storage (S3, GCS, ADLS). Flexible: raw data сохраняется как есть. ELT: load first, transform later. Ideal для data science, ML training, exploratory analysis.

Lakehouse

Best of both worlds. Data lake storage + warehouse features (ACID, indexes, SQL). Delta Lake, Apache Iceberg, Apache Hudi — three main formats. Databricks, Dremio, Snowflake ecosystems. Modern architecture 2024–2025 для новых платформ.

iGaming choice

Real-time analytics: streaming warehouse (ClickHouse, StarRocks). ML training: data lake для raw events. Regulatory reporting: SQL warehouse с strong ACID. Часто hybrid: raw events в lake, aggregated в warehouse. Cost optimization: hot data в warehouse, cold в lake.