MatrixOne v2.0.0 Release Note
We are excited to announce the official release of MatrixOne Core v2.0.0 on October 31, 2024!
MatrixOne is an AI-driven, cloud-native hyper-converged database that leverages a storage-compute separation architecture, fully utilizing cloud infrastructure. It is MySQL-compatible and supports hybrid workload scenarios, combining vector data types and full-text search capabilities to efficiently handle multimodal data queries and management for generative AI applications.
MatrixOne v2.0.0 Feature Overview
In this release, MatrixOne has significantly enhanced support for generative AI applications, disaster recovery, and stability. Key features include external storage access, unstructured data support, full-text search capabilities, improved vector search performance, snapshot backups, point-in-time recovery (PiTR), CDC, and log-based primary-secondary cluster disaster recovery. Compatibility with MySQL has also been further improved. With these features, MatrixOne is becoming an ideal choice for enterprises looking to build AI-driven intelligent data management platforms.
Use Cases for MatrixOne
MatrixOne is suitable for the following application scenarios. We invite users with these business needs to contact us for trial testing.
Generative AI Scenarios: MatrixOne provides robust multimodal data support, real-time retrieval, and intelligent data processing for generative AI, serving as core infrastructure for applications such as text and image generation. It offers high-efficiency data management, vector and hybrid retrieval, Python UDF data cleaning and preprocessing, and GPU-accelerated real-time inference capabilities. MatrixOne supports high-volume data access and storage, online inference, and dynamic feedback, making it ideal for the rapid deployment, iteration, and optimization of generative AI applications in enterprise environments.
Time-Series Data Applications: In modern IoT applications, billions of devices and sensors continuously collect and transmit data, producing terabytes of real-time data daily. MatrixOne's hyper-converged database offers high-performance real-time data processing, with millisecond-level high-concurrency writes, fast retrieval, and excellent scalability to handle peak loads. It integrates seamlessly with machine learning models for predictive maintenance, efficiency optimization, and smart monitoring, addressing the demands for high throughput, low latency, and intelligent data management in IoT applications.
Hybrid Workload Support: In traditional business systems such as OA, ERP, and CRM, data growth and business complexity often strain the performance of standalone databases. MatrixOne's hybrid workload support enables enterprises to meet both operational and analytical demands within a single database, eliminating the need for additional systems. With real-time analytics, it provides quick responses under high concurrency, allowing seamless scaling and maintaining efficiency even as data volume increases.
Enterprise SaaS Scenarios: As enterprise SaaS applications grow, the need to support multi-tenant models is essential. MatrixOne's multi-tenant architecture offers load isolation, independent scaling, and unified management, reducing management costs while ensuring data isolation and operational efficiency. MatrixOne is an ideal choice for SaaS applications requiring cost control, ease of management, and data isolation.
Key New Features in MatrixOne v2.0.0
Multimodal Data Management: MatrixOne supports direct access to external storage, remote file systems, and local file systems through Stage objects and direct access to storage system files via the datalink type. This is particularly useful in generative AI applications, enhancing development efficiency and reducing operational costs.
Full-Text Indexing for Text or JSON Data: Full-text indexing on JSON or TEXT columns improves performance in AIoT applications. Combined with JSON data types, this reduces data redundancy, enhancing MatrixOne's competitiveness in AIoT scenarios.
Vector Search Optimization: This update optimizes vector search performance, enabling efficient vector distance-based retrieval for large datasets, essential for LLM and RAG-based generative AI applications.
Snapshot-Based Backup and Recovery: Cluster or tenant data snapshots allow fast state capture at specific points and ensure rapid recovery in case of failure. The feature supports cross-tenant recovery, enhancing disaster recovery capabilities.
Log-Based Primary-Secondary Cluster Disaster Recovery: Transaction log synchronization ensures high availability and disaster recovery. When the primary database fails, the secondary quickly takes over to maintain uninterrupted service.
Point-in-Time Recovery (PiTR): Capturing all data changes after an initial snapshot, PiTR allows database restoration to a precise historical point, minimizing storage requirements and enhancing recovery efficiency.
MatrixOne to MySQL CDC: Real-time synchronization from MatrixOne to MySQL enables disaster recovery and maintains a backup link for users migrating from MySQL to MatrixOne.
Table-Level Publish-Subscribe Functionality: In this iteration, we introduced table-level publish-subscribe for more granular data synchronization, allowing real-time updates for specific tables without exposing other data.
Additional New Features
SQL Statements
Support for rename table
Support for create pitr, drop pitr, alter pitr, restore pitr, and show pitrs
Optimizations for show publications and show subscriptions
load data infile now supports loading in user-defined column order
Data Types
Support for datalink data type
Indexes and Constraints
Full-text indexing support
Functions and Operators
Functions for JSON data type, including json_row, jq, try_jq, json_extract_string, and json_extract_float64
Arithmetic operations on the date returned by now()
Tools
mo-backup: manages PiTR tasks
mo_cdc: manages CDC tasks
MySQL Compatibility
Case-insensitive search in where clauses for Select statements
Support for Encode()/Decode() functions
Quick Start
Community users and enterprise developers can deploy MatrixOne with a single command for trial testing.
Our documentation website provides detailed architecture explanations, installation guides, and development tutorials to help you explore MatrixOne's capabilities. Additionally, our GitHub page and community groups welcome questions, discussions, and feedback.
Known Issues
Disaster recovery for primary-secondary clusters does not support external tables or data in stages.
Secondary clusters in disaster recovery only support cold backup and cannot be opened in read-only mode.
CDC supports only table-level data synchronization.
Snapshot backups currently support only cluster and tenant-level backups but can restore to cluster, tenant, database, or table level.
Snapshots and PiTR cannot restore deleted tenant data.