A Massively Multi-user Datastore, Synced with Mobile Clients

Summary

The article discusses the architectural design of a massively multi-user datastore developed at Square, which is tailored to manage extensive merchant catalogs synced with mobile clients. It emphasizes the challenges of maintaining data consistency and availability across various client states, particularly when offline. The solution employs an entity-attribute-value model to provide both structure and flexibility, allowing for rapid feature development and custom data integration. Key features include synchronizable constraints, an append-only data model for transactional integrity, and a rollback mechanism for maintaining historical data states. The design aims to ensure that clients can operate with predictable data formats while allowing for extensibility and efficient data management.

Key Learnings

• 1The use of an entity-attribute-value model allows for a flexible yet structured approach to data management, accommodating diverse client needs.
• 2Implementing synchronizable constraints as catalog objects enhances data integrity and client-side validation without requiring schema changes.
• 3An append-only data model facilitates transactional properties while enabling efficient rollback and history tracking capabilities.
• 4The architecture supports both location-aware and merchant-aware clients, optimizing data delivery based on client context.
• 5By allowing clients to create new object types and attribute definitions via API, the system promotes rapid iteration and customization.

Who Should Read This

Senior Software Architects designing scalable, multi-tenant data solutions for mobile applications.

Test Your Knowledge

Topics