Choosing an Azure Data Platform: PostgreSQL, MySQL, Redis, Synapse, Data Lake, Data Factory and Fabric
Open Source Databases
- Azure Database for PostgreSQL – Flexible Server and MySQL – Flexible Server: managed, with zone-redundant HA, automated backups, read replicas, VNet integration or private endpoint, and stop/start to save cost
- Use them to move open-source apps to the cloud with little change
bash
az postgres flexible-server create -g rg-data -n pg-shop-482 -l centralindia --tier GeneralPurpose --sku-name Standard_D2ds_v5 \
--version 16 --high-availability ZoneRedundant --storage-size 128 --admin-user pgadmin --admin-password '<StrongPassword>'
Azure Cache for Redis
- A managed in-memory cache for sessions, leaderboards, rate limits and caching database results
- Pattern: cache-aside. Read from cache, on a miss read the database and fill the cache with a TTL
- Premium and Enterprise tiers add clustering, persistence, zone redundancy and geo-replication
- Check the current offering (Azure Managed Redis is the newer option) before deciding
Analytics Services
| Service | What it is |
|---|---|
| Azure Data Lake Storage Gen2 | Cheap, scalable lake on Blob with hierarchical namespace |
| Azure Data Factory | Orchestrate and move data (ETL/ELT) with pipelines and 90+ connectors |
| Azure Synapse Analytics | Data warehouse (dedicated SQL pool), serverless SQL, Spark, pipelines |
| Microsoft Fabric | SaaS analytics platform: OneLake, lakehouse, warehouse, pipelines and Power BI in one product |
| Azure Databricks | Apache Spark based platform for data engineering and machine learning |
| Azure Stream Analytics / Event Hubs | Real-time streaming ingestion and queries |
| Azure Data Explorer | Fast analytics on logs and time-series |
| Power BI | Dashboards and reports |
Note Microsoft is investing heavily in Fabric. Check the current guidance for new analytics projects, because product direction changes.
The Big Decision Table
| Requirement | Service |
|---|---|
| Relational app, minimal admin | Azure SQL Database |
| SQL Server with Agent and linked servers | SQL Managed Instance |
| Open-source relational | PostgreSQL / MySQL Flexible Server |
| Global JSON documents, millisecond latency | Cosmos DB (NoSQL) |
| Cache and sessions | Redis |
| Images, files, backups | Blob Storage |
| Big data lake | Data Lake Storage Gen2 |
| Enterprise data warehouse | Synapse dedicated SQL pool or Fabric Warehouse |
| Streaming events | Event Hubs + Stream Analytics |
| Simple key-value at low cost | Table Storage |
| Full-text search | Azure AI Search |
| Graph relationships | Cosmos DB Gremlin |
Reference Architecture: Modern Data Platform
- Sources (apps, databases, files) are ingested by Data Factory or Event Hubs
- Raw data lands in Data Lake Storage (bronze), then is cleaned (silver) and modeled (gold) with Spark or SQL
- A warehouse or lakehouse serves reports
- Power BI shows dashboards
- Purview catalogs and governs data
Security for Data Services
- Use private endpoints, Entra ID auth and managed identities
- Use customer-managed keys where policy requires them
- Turn on auditing and Defender plans
- Mask or classify sensitive columns
Practice
- Choose a service for each row of the decision table without looking
- Draw the cache-aside flow for a product catalog
- Explain why Fabric and Synapse may both appear in an architecture
Conclusion
Pick the simplest managed service that meets the requirement: SQL for relational, Cosmos DB for global NoSQL, Redis for cache, Blob and Data Lake for files and big data, and Data Factory, Synapse or Fabric for analytics.
Quick Summary: open source databases; Redis; analytics; decision table; reference architecture.

