Azure SQL Hyperscale: 100TB Databases

Traditional Azure SQL Database is limited by the disk size of the underlying VM (4TB max usually). **Hyperscale** decouples compute from storage, allowing databases to grow up to 100TB with rapid scaling.

Architecture

flowchart TB
    App --> Compute[Primary Compute Node]
    Compute -->|Log Service| Log[Log Service]
    Log -->|Async| Page1[Page Server 1]
    Log -->|Async| Page2[Page Server 2]
    Page1 --> Storage[Remote Storage (Azure Blob)]
    Page2 --> Storage
    
    style Compute fill:#E1F5FE,stroke:#0277BD

Why use it?

• **Instant Backups**: Backups are file-snapshots, taking zero IOPS from the compute.
• **Fast Restore**: Restoring a 50TB database takes minutes, not days.
• **Read Replicas**: You can add 0-4 replicas for read scale-out dynamically.

Key Takeaways

• Hyperscale is the future tier for general purpose workloads.
• Migration is one-way (you can’t go back to General Purpose easily).