Azure SQL Hyperscale Deep Dive
summary
Architecture explains that Azure SQL Hyperscale uses separation of compute and storage with stateless compute, an externalized log service (durable landing zone with ~2–4 ms commit latency and 7‑day retention) and page servers with 128GB files backed by blob storage; compute holds RBPEX and tempdb on local SSD. Reads scale out to up to four read replicas, applied asynchronously, which in a demo doubles throughput (≈2,100 → ≈4,400 tps). Migrations from on‑prem are size‑of‑data moves or bulk loads; index builds should use resumable or MAXDOP limits.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 0:46What is Hyperscale
- 4:28Near realtime
- 5:46Data marts
- 7:14Architecture
- 10:52Log Service
- 13:22Paid Server
- 16:51Readable Secondary
- 20:24Azure Standard Storage
- 22:57Demo
- 23:08Workload Driver
- 24:14Running a workload
- 32:11Best practices troubleshooting
- 40:35Supportability diagnostics
- 45:36Migration
related talks
49mBuild & scale applications with Azure SQL Hyperscale
17mAutomatic compute scaling with serverless for Hyperscale in Azure SQL Databases
52mMore for less: Cost optimizing your Azure SQL databases
01h 15mChoose Azure SQL Database and Learn What Every Developer Should Know
48mBuilding SaaS solutions on Azure SQL Database
54mAzure Best Practices for Lift and Shift
47mPerfecting business continuity for Azure SQL DB
20mChoosing the Right Data Store--An Overview of Azure Data Platform Choices