Group 6: Relational Data

Relational managed offerings: Azure SQL Database, Azure SQL Managed Instance, SQL Server on VM, PostgreSQL, MySQL, MariaDB, plus Database Migration Service (transition helper). Differences revolve around control surface, feature coverage, scaling model, and migration friction.

Rule of Thumb: Prefer managed PaaS (Azure SQL DB / SQL MI / OSS managed) unless you have feature gaps or OS-level dependencies requiring SQL on VM.

Services & Core Identity

Azure SQL DB

Fully managed single DB / elastic pools. Automatic patching, built-in HA. Limited surface vs full instance.

Use: new cloud-native workloads.
Scaling: DTU or vCore, serverless tier.
Features: Hyperscale for large sizes.

SQL Managed Instance

Near full SQL Server surface (Agent, cross-db) with PaaS ops. Instance-level features restored.

Use: lift from on-prem w/ minimal change.
Managed backups, patching.
VNet integration.

SQL on VM

IaaS-hosted SQL. Full OS + instance control; manual patching/HA unless configured.

Use: unsupported features, custom agents.
Higher ops overhead.
Can use Azure Backup.

PostgreSQL

Managed OSS relational (Flexible Server). Extensions support; versioned upgrades.

Use: open-source stack preference.
High concurrency, rich types.
Can scale compute/storage independently.

MySQL

Managed MySQL (Flexible). Popular LAMP stack DB; read replicas, zone redundant HA.

Use: web apps relying on MySQL dialect.
Per-second billing on flexible.
Performance tiers (Burst / Gen Purpose / Memory).

MariaDB

Managed MariaDB (legacy workloads). Similar to MySQL; may have different engine optimizations.

Use: existing MariaDB apps.
Consider MySQL/PostgreSQL for new builds.
Migration path via DMS.

DMS

Database Migration Service orchestrates minimal downtime migrations and assessments.

Use: shift SQL/OSS DB to managed.
Helps with schema + data movement.
Not a runtime target.

Key Differences

Dimension	SQL DB	SQL MI	SQL VM	PostgreSQL	MySQL	MariaDB
Control Surface	Database only	Instance-level	Full OS/Instance	Instance-level	Instance-level	Instance-level
HA Built-in	Yes	Yes	Manual config	Yes	Yes	Yes
Scaling Granularity	Per DB/pool	Per instance	VM size	Compute/Storage	Compute/Storage	Compute/Storage
Serverless Option	Yes (auto-pause)	No	No	No	No	No
Feature Parity w/ Full SQL	Subset	Near-full	Full	N/A	N/A	N/A
Extensions	Limited	Limited	Full (OS install)	Rich (Postgres ext)	Plugins	Plugins
Ops Overhead	Lowest	Low	High	Low	Low	Low
Ideal Use	Cloud-native app	Lift & shift SQL	Legacy / full control	OSS structured data	LAMP style apps	Existing MariaDB

Mathematical Selection Model (Runtime)

Models relative fit. Criteria [0..10]. SQL VM and DMS included but DMS always flagged as helper only.

Score_SQLDB = 0.30*C_cloud + 0.20*C_lowops + 0.15*C_autoScale + 0.15*(10 - C_fullSurface) + 0.10*(10 - C_ext) + 0.10*C_serverless Score_SQLMI = 0.30*C_instance + 0.20*C_feature + 0.15*C_network + 0.15*C_lowops + 0.10*(10 - C_serverless) + 0.10*(10 - C_ext) Score_SQLVM = 0.30*C_fullSurface + 0.20*C_ext + 0.15*C_osControl + 0.15*C_feature + 0.10*(10 - C_lowops) + 0.10*C_license Score_PG = 0.35*C_ext + 0.20*C_cloud + 0.15*C_open + 0.10*C_lowops + 0.10*C_scalability + 0.10*C_json Score_MySQL = 0.30*C_lamp + 0.20*C_cloud + 0.15*C_lowops + 0.10*C_scalability + 0.10*(10 - C_fullSurface) + 0.15*C_replica Score_Maria = 0.30*C_existing + 0.20*C_lowops + 0.15*C_cloud + 0.10*C_scalability + 0.10*C_ext + 0.15*(10 - C_migrate)

SQL DB {{vm.scores.sqldb | number:2}}

SQL MI {{vm.scores.sqlmi | number:2}}

SQL VM {{vm.scores.sqlvm | number:2}}

PostgreSQL {{vm.scores.pg | number:2}}

MySQL {{vm.scores.mysql | number:2}}

MariaDB {{vm.scores.maria | number:2}}

Recommended: {{vm.recommended.name}} ({{vm.recommended.score | number:2}})

Interpretation Rules

If C_fullSurface ≥ 7 OR C_osControl ≥ 6 ⇒ SQL VM (verify TCO).
If C_instance ≥ 6 AND C_feature ≥ 6 ⇒ SQL MI.
If C_serverless ≥ 6 OR unpredictable load ⇒ SQL DB (serverless tier).
If C_open ≥ 6 AND C_ext ≥ 6 ⇒ PostgreSQL.
If C_lamp ≥ 6 ⇒ MySQL (or Maria if existing base).

When NOT to Use Relational

Massive scale, low latency key-value ⇒ Cosmos DB / Redis.
Semi-structured analytics at scale ⇒ Data Lake / Synapse.
Write path > vertical scaling limit ⇒ consider sharding or alt model.

Summary

SQL DB for most new SQL workloads; SQL MI for near full SQL features with less admin; SQL on VM only for OS/feature gaps; PostgreSQL for extensibility & rich types; MySQL/MariaDB for existing ecosystems. Start with ops burden & feature parity, refine via scaling + ecosystem needs.