Azure Database for PostgreSQL

Overview

Azure Database for PostgreSQL is Microsoft Azure’s fully managed, native PostgreSQL database service. It’s the direct equivalent to self-hosted PostgreSQL, providing enterprise-grade features with minimal operational overhead.

Service Tiers

1. Flexible Server (Recommended - Currently Used)

  • What it is: Modern deployment option with maximum control and flexibility
  • Best for: Production workloads, development, and testing
  • Features:
    • Zone-redundant high availability (99.99% SLA)
    • Independent compute and storage scaling
    • Better cost optimization with stop/start capability
    • More control over maintenance windows
    • Support for PostgreSQL 11, 12, 13, 14, 15, 16

2. Single Server (Legacy)

  • What it is: Original managed PostgreSQL offering
  • Status: Being phased out in favor of Flexible Server
  • Best for: Simple applications (not recommended for new deployments)

Current Implementation

The Engram platform uses Azure Database for PostgreSQL Flexible Server as configured in infra/main.bicep:

resource postgres 'Microsoft.DBforPostgreSQL/flexibleServers@2021-06-01' = {
  name: '${envName}-db'
  sku: {
    name: 'Standard_B1ms'  // Burstable tier - cost-optimized
    tier: 'Burstable'
  }
  properties: {
    version: '13'
    storage: {
      storageSizeGB: 32
    }
  }
}

Key Features

1. AI Integration

  • Native vector search: Built-in support for AI/ML workloads
  • Azure AI extension: Direct integration with Azure AI services
  • Local embeddings: Process embeddings within the database

2. High Availability

  • 99.99% uptime SLA with zone-redundant configuration
  • Automatic failover for mission-critical applications
  • Built-in backup and point-in-time restore

3. Cost Optimization

  • Scale-to-zero: Stop server when not in use (dev/test)
  • Burstable SKUs: Cost-effective for variable workloads
  • Up to 58% cost reduction vs on-premises

4. Security & Compliance

  • Azure AD authentication: Integrated with Entra ID
  • Private endpoints: Isolated network connectivity
  • Encryption at rest: Automatic data encryption
  • Compliance: SOC, ISO, HIPAA, PCI-DSS certified

5. PostgreSQL Compatibility

  • Latest versions: Support for PostgreSQL 11-16
  • Extensions: Full support for PostgreSQL extensions
  • Migration: Easy migration from on-premises PostgreSQL

SKU Tiers

Burstable (B-series) - Current Choice

  • Best for: Development, testing, low-traffic applications
  • Examples: Standard_B1ms, Standard_B2s
  • Cost: ~$12-25/month (varies by region)
  • Features: Can burst CPU when needed

General Purpose

  • Best for: Most production workloads
  • Examples: Standard_D2s_v3, Standard_D4s_v3
  • Cost: ~$100-400/month
  • Features: Predictable performance, balanced compute/memory

Memory Optimized

  • Best for: High-performance, memory-intensive workloads
  • Examples: Standard_E2s_v3, Standard_E4s_v3
  • Cost: ~$200-800/month
  • Features: High memory-to-vCPU ratio

Comparison with Alternatives

Service Use Case Pros Cons
Azure Database for PostgreSQL PostgreSQL workloads Native PostgreSQL, managed, AI-ready PostgreSQL-specific
Azure SQL Database SQL Server workloads Microsoft ecosystem, advanced features Not PostgreSQL
Azure Cosmos DB NoSQL, global scale Multi-model, global distribution Different data model
Self-hosted PostgreSQL on VM Full control Complete control Operational overhead

Connection Configuration

Current Setup (from backend/core/config.py) 

postgres_host: str = Field("localhost", alias="POSTGRES_HOST")
postgres_port: int = Field(5432, alias="POSTGRES_PORT")
postgres_user: str = Field("postgres", alias="POSTGRES_USER")
postgres_password: str = Field("password", alias="POSTGRES_PASSWORD")
postgres_db: str = Field("engram", alias="POSTGRES_DB")

Azure Flexible Server Connection String 

postgresql://<admin-user>:<password>@<server-name>.postgres.database.azure.com:5432/<database-name>?sslmode=require

Best Practices

  1. Use Flexible Server for all new deployments
  2. Enable private endpoints for production (network isolation)
  3. Use Azure AD authentication when possible (better security)
  4. Configure automated backups (7-35 day retention)
  5. Monitor with Azure Monitor (built-in metrics and alerts)
  6. Use connection pooling (PgBouncer recommended)
  7. Scale compute independently from storage

Cost Optimization Tips

  1. Use Burstable SKUs for dev/test (can stop/start)
  2. Right-size storage (32GB minimum, pay for what you use)
  3. Enable auto-grow to avoid over-provisioning
  4. Use reserved capacity for predictable workloads (up to 55% savings)
  5. Stop servers during non-business hours (dev/test)

Migration Path

From Self-Hosted PostgreSQL

  1. Use pg_dump to export data
  2. Create Azure Database for PostgreSQL Flexible Server
  3. Use pg_restore to import data
  4. Update connection strings
  5. Test and validate

From Single Server (Legacy)

  1. Use Azure Database Migration Service
  2. Or manual migration with pg_dump/pg_restore
  3. Update application connection strings

Resources

Summary

Azure Database for PostgreSQL Flexible Server is the native, fully managed PostgreSQL service on Azure. It provides:

✅ 100% PostgreSQL compatibility
✅ Enterprise-grade availability (99.99% SLA)
✅ AI-ready features (vector search, embeddings)
✅ Cost-effective (scale-to-zero, burstable SKUs)
✅ Secure by default (encryption, private endpoints, Azure AD)
✅ Fully managed (patching, backups, monitoring)

The Engram platform is already configured to use this service, making it the ideal choice for production deployments.