Scalable Applications Without Polyglot tax: Azure SQL Hyperscale | OD824

Microsoft Developer explains how Azure SQL Database Hyperscale can reduce polyglot persistence complexity by consolidating multiple workload patterns into a single SQL platform, covering multi-model capabilities, transactional guarantees, scaling options, and operational features like snapshot backups and readable replicas.

Overview

Modern application architectures often adopt polyglot persistence (multiple specialized databases), which can increase:

• System complexity
• Data movement between stores
• Operational overhead

This session presents Azure SQL Database Hyperscale as an alternative: a single, multi-model SQL platform intended to support relational and transactional workloads while also addressing emerging needs (including JSON-centric patterns) within SQL.

What the session covers

Why polyglot persistence happens

• Adding new application features can push teams toward introducing additional, specialized databases.
• Over time, this can create a fragmented data layer with more moving parts to operate and integrate.

Multi-model and schema flexibility with JSON

• Hyperscale is positioned as supporting schema flexibility and transactional support for JSON operations within SQL.
• The goal is to handle document-style patterns without moving data into a separate document database.

Performance positioning vs document databases

• The session includes a performance comparison message against document databases.
• The core argument is that consolidating on a single platform can be a simpler path while still meeting performance needs.

Transactional guarantees for agent-driven systems

• The session highlights the importance of:
  • ACID semantics
  • Idempotency
• These are framed as especially important in “agent-driven” systems where repeated or retried operations must remain correct.

Operational reliability: snapshot backups

• Azure snapshot backups are presented as a way to:
  • Improve reliability
  • Reduce compute load associated with backup operations

Scaling patterns: primary OLTP and readable replicas

• Hyperscale usage is discussed in terms of:
  • A primary OLTP database
  • Named readable replicas for read scale-out scenarios

Engine and platform notes

• The session calls out the integration of the Microsoft SQL core engine in Hyperscale.
• It also mentions a licensing-free model message in the context of Hyperscale.

Resources

• https://aka.ms/build26-next-steps

Session metadata

• Event: Microsoft Build 2026
• Session: OD824
• Level: Intermediate
• Track: Cloud platform & data
• Language: English (US)

Chapters

• 0:00 - Introduction and session overview
• 00:01:05 - Adding new features leads to multiple specialized databases
• 00:07:01 - Schema flexibility and transactional support for JSON operations
• 00:08:03 - Performance comparison with document databases and single platform message
• 00:13:16 - Importance of ACID semantics and idempotency in agent-driven systems
• 00:16:11 - Azure snapshot backups improve reliability and reduce compute load
• 00:17:02 - Using hyperscale with primary OLTP and named readable replicas
• 00:19:19 - Integration of Microsoft SQL core engine in Hyperscale
• 00:22:42 - Performance and licensing-free model of Azure SQL Hyperscale