Product Overview

Technical Deep Dive

News

Whitepaper

Use Case

Access authoritative materials on MatrixOne Intelligence products, cutting-edge AI technologies, and industry best practices to drive your future development.

Resources

MatrixOne Intelligence 3.2 Release: Unlock a New Experience in Multimodal Data Intelligence

MatrixOne Intelligence AI Native Multimodal Data Intelligence Solution White Paper

Announcing: "MatrixOne Intelligence AI Native Multimodal Data Intelligence Solution White Paper" – Click to Download for Free!

BeeDa adopts MatrixOne Cloud Serverless as its sole database for ERP system, achieving cost reduction with performance improvement and business isolation.

BeeDa Software -- A Leading Software Brand in the Wallpaper Industry

MatrixOne Intelligence powers an AI bidding document agent that automates end-to-end bid preparation, reducing average daily preparation time from hours to minutes with unified multi-modal data governance.

Intelligent Upgrading and Transformation Project: AI Bidding Document

ETAO Innovation adopts MatrixOne to replace MySQL, InfluxDB, and MongoDB, achieving a unified MES system architecture that significantly improves development efficiency and data processing performance.

Shenzhen ETAO Innovation | MatrixOne Empowers ETAO Innovation to Build a High-Performance Intelligent Manufacturing AIoT System

Extreme Vision adopts MatrixOne to build a unified AI feature data platform, reducing model development cycle by 30% and improving cross-department data utilization by 40%.

Extreme Vision | MatrixOne Empowers Extreme Vision to Build a Unified and Reusable AI Feature Platform

Guangheng adopts MatrixOne Serverless to replace MySQL for tile inspection data, achieving order-of-magnitude improvements in IoT data ingestion and query performance with automatic scaling.

Guangheng | Expert in Inspecting Complex Industrial Defects with MatrixOne

A global health supplement retail brand adopts MatrixOne Intelligence to build a unified AI data engineering platform, reducing data preparation time by 80% and enabling rapid AI application deployment.

A Well-known Health Supplement Enterprise: Introducing MatrixOne Intelligence, Building a Unified GenAI Data Platform for Sales Copilot

A Grade III Class A hospital in China leverages MatrixOne Intelligence to build an IBS intelligent diagnosis system, transforming clinical data assets into AI-ready resources for enhanced diagnosis and treatment.

A Grade-A Hospital: Building an IBS Intelligent Diagnosis System Based on MatrixOne Intelligence

Jiangxi Copper Group adopts MatrixOne to build an AIoT smart operations system for flash smelting, shortening implementation cycle by 60% and reducing copper accumulation at furnace bottom by 80%.

Jiangxi Copper Group | MatrixOne Empowers the Development of an AIoT Smart Operations System

Shenzhen Smart City adopts MatrixOne to transform its traffic big data platform, achieving second-level real-time analytics, reducing system complexity by 80%, and cutting operation costs by 50%.

Shenzhen Smart City | One-stop Traffic Big Data Platform Transformation Based on the Hyper-Converged Database MatrixOne

Skyable adopts MatrixOne to consolidate MySQL, MongoDB, and Elasticsearch into a single platform, significantly simplifying architecture and improving IoT data processing efficiency.

Skyable | MatrixOne Empowers Skyable to Build an Integrated Internet of Things Platform

StoneCastle adopts MatrixOne HTAP database to replace MySQL, reducing complex aggregation query times from 10-30 minutes to 5-10 seconds while maintaining private deployment and high MySQL compatibility.

StoneCastle | MatrixOne Helps StoneCastle Build an Efficient Financial Analytics System

Suwen Intelligence adopts MatrixOne to simplify its complex multi-database architecture into a unified platform, reducing delivery time from two months to one week and enabling minute-level data processing.

TechAgent of Shenzhen Suwen Intelligence | MatrixOne Builds a Hyper-Converged AIGC Data Infrastructure Platform

A telecommunications operator adopts MatrixOne cloud-native HTAP solution for connected vehicle management, reducing hardware resources by 80% and improving query performance by 20x.

A Telecommunications Operator | MatrixOne Builds a Low-Cost, High-Performance Connected Vehicle (IoV) Management System

A tile manufacturing enterprise adopts MatrixOne Intelligence to build TaoXin, an intelligent image-based search platform enabling sales personnel to quickly locate products through photos, images, or text.

A Large Tile Manufacturing Enterprise | MatrixOne Intelligence Empowering the Construction of an Intelligent Tile Image Search Platform

Uascent Technology leverages MatrixOne and MatrixGenesis to build an intelligent voice assistant with natural language understanding, enabling users to control smart home appliances through natural language interaction.

A Globally Leading AIoT Development and Cloud Service Platform

UNRE Technology adopts MatrixOne Cloud Serverless to replace RDS MySQL, achieving multi-tenant isolation, improved performance, and nearly 50% cost reduction for construction measurement SaaS platform.

UNRE Technology | MatrixOne Empowers UNRE Technology to Build Construction Informatization Solutions

In the IoT era, Skyable and MatrixOne have partnered to build a next-gen platform, enhancing device connectivity, data processing, and intelligent forecasting while simplifying system architecture.

MatrixOne Empowers Skyable to Replace MySQL, MongoDB, and Elasticsearch with a Unified IoT Platform

Matrix Origin and 21Vianet are co-developing an AI data intelligence and hyper-connectivity computing platform. CEO Wang Long highlighted data governance and large model challenges, advancing intelligent computing and digital transformation for enterprises.

Matrix Origin CEO Wang Long Attends the Launch Event of the Hyper-Connectivity (Suzhou) Headquarters Node on 1024

StoneCastle, a U.S. financial service provider, faced challenges with their MySQL accounting management system due to rapid data growth and query performance bottlenecks. To enhance efficiency and ensure data security, they migrated to MatrixOne, a high-performance HTAP database. This transition improved data processing, reduced query times significantly, and allowed for secure private cloud deployment while maintaining compatibility with existing MySQL systems.

MatrixOne Helps StoneCastle Build a High-Performance Financial Analysis System

ComfyUI is an open-source image generation tool with GPU support. Neolink.AI simplifies its deployment by providing powerful GPU resources and pre-configured environments, easing hardware setup and optimization.

ComfyUI has now launched Neolink.AI, making AI drawing easy to master!

Jiangxi Copper, a top global copper producer, implemented an AIoT solution at its Guixi Smelter to automate copper and slag discharge. Using MatrixOne, this project reduced copper in slag by 80%, improved production precision, and shortened implementation time by 60%, driving data-driven automation in smelting operations.

MatrixOne Helps Jiangxi Copper Corporation Limited Build an AIoT Big Data System for Smart Operations at the Furnace Front

ETAO Innovation simplified its MES system by switching to MatrixOne—no more juggling MySQL, InfluxDB, or MongoDB! Now, they handle everything with one database, improving performance and speeding up project delivery.

MatrixOne Assists ETAO Innovation in Building a High-Performance Smart Manufacturing AIOT System

A telecom operator revamped its connected vehicle management using MatrixOne’s cloud-native HTAP solution. With 20x faster queries and dynamic scalability, they boosted performance and cut costs—all while simplifying operations.

MatrixOne Helps a Telecom Operator Build a Low-Cost, High-Performance Connected Vehicle Management System

ShenZhiCheng is revolutionizing its traffic platform using MatrixOne’s hyper-converged database. Simplified architecture, seamless cloud-native integration, and real-time processing—this transformation is a game-changer!

ShenZhen Smart City’s One-Stop Traffic Big Data Platform Transformation Based on the Hyper-Converged Database MatrixOne

We showcased our groundbreaking MatrixOne database and explored AI-driven data management at the 15th China Database Technology Conference. Discover how we're shaping the future of databases!

Event Review: Matrix Origin at the 15th China Database Technology Conference (DTCC 2024)

Understanding MOserver-MySQL client communication, MySQL protocol structure (Packet Length, Sequence ID, Payload), encoding types (fixed/variable length integers, strings).

MatrixOne's MySQL Protocol Implementation

Matrix Origin captivated WAIC 2024 with MatrixOS, their AI-Native Data Intelligent OS. Recognized as a "FUTURE TECH100—Innovation Partner," they are driving digital transformation.

MatrixOrigin Successfully Concludes WAIC 2024

Dive into the highlights of "Open Source Summer 2024" with MatrixOne! This year, 3 innovative projects shone through, led by talented students under the guidance of expert mentors. Congratulations!

Congratulations to MatrixOne Open Source Community Project Participants

MatrixOrigin aced ISO audits (9001, 20000, 27001) A testament to our commitment to quality and excellence, setting the stage for ongoing improvement and customer value.

ISO Annual Audit Passed

Readers familiar with database distributed transactions should understand that debugging bugs related to the correctness of distributed transactions is a very challenging task. This article mainly introduces how MatrixOne debugs transaction correctness issues during its development process.

Efficiently Debugging Transaction Correctness Bugs

The former VP of Tencent Cloud, Mr. Zhang Shuxiang, has joined MatrixOrigin as a Partner and Senior VP.

Former Tencent Cloud Vice President Zhang Shuxiang Joins MatrixOrigin to Co-create a New Chapter in Artificial Intelligence

MatrixOne upgrading to MatrixOS is a "response" to market demand under the wave of AI.InfoQ conducted a special interview with Wang Long, CEO of MatrixOrigin, focusing on the business expansion plans of MatrixOrigin and discussing the future development of databases and AI technology.

MatrixOne → MatrixOS: "AI Infra" and "AI Platform" Rewrite the Entrepreneurial History of MatrixOrigin

on May 22, GeekPark's Tech Talk Tonight invited Wang Long, founder and CEO of MatrixOrigin, to discuss the foundational elements of large model development:"data and computing power."

Without "Operating System", AI Doesn't Work Well Even If It's Free

Wang Long, the founder and CEO of Matrix Origin, was invited to attend the GIIC - 2024 GPU Infrastructure Innovation conference and delivered a keynote speech on the topic of "Computing Power + Data The Cornerstone of the AI Native World".

MatrixOrigin CEO Wang Long Delivers Keynote Speech at GIIC Conference

MatrixOne is a distributed hyper-converged heterogeneous database. The key functions of this iterative update include snapshot backup, snapshot backup, incremental physical backup, CTAS, BITMAP, and vector index. With the introduction of these new functions, MatrixOne is gradually becoming an ideal choice for enterprise-level data management and analysis.

MatrixOne Core Version 1.2.0 Officially Released

To address the challenges of the AI Big Model era, Matrix Origin has launched the MatrixOS product. MatrixOS consists of three core sub-products, which can either be combined into an all-in-one solution or independently deployed to provide services.

Building an Open-Source AI Native Software Platform Around Computational Power & Data by MatrixOrigin

Apache Spark is a distributed computing engine designed for efficient processing of large-scale data. It employs distributed parallel computing, distributing tasks of data splitting, computing, and merging across multiple computers, thus achieving efficient data processing and analysis.

How to Use Spark to Write Batch Data into MatrixOne

The execution plan for shuffle is a very important part of the optimizer. Due to space limitations, only some key aspects have been introduced here. For more details and related implementation code, feel free to directly check the MatrixOne source code.

Advanced Shuffle Optimization Techniques — Join reorder

Previously, the tpch dataset was always used for examples, but tpch represents an ideal scenario where all data are evenly distributed. In actual production environments, many datasets are unevenly distributed. For uneven data, a straightforward approach is to use a hash shuffle to ensure that the data is evenly distributed after bucketing.

Advanced Shuffle Optimization Techniques：How do we handle uneven data?

Although the execution plan of shuffle on multiple CNs can reduce the overhead of hash tables, it may also increase the cost of data transmission over the network. A good shuffle execution plan must minimize data transmission over the network as much as possible. Therefore, the introduction of colocate shuffle optimization is crucial for performance.

Introduction to Colocate Shuffle

The basic principle of the shuffle algorithm is to bucket the input data, where each bucket's data can be processed independently and output results. This reduces the size of the hash table when aggregating data.

Why is shuffle support needed?

We will start with a simple introduction of the structure of a database kernel.

Quickly Start: MatrixOne Database Kernel Front-End

BPF technology has been listed as one of the hottest new areas in the Linux kernel domain in recent years. It has successfully endowed the Linux kernel with a degree of dynamic programmability, allowing for real-time modifications to the kernel's behavior during runtime, without the need to recompile or reboot the kernel.

BPF Development: Starting with Hello World

In simple terms, the main content of the Logtail protocol is divided into two parts:in-memory data and metadata. The core difference lies in whether the data has been transferred to object storage.

MatrixOne Logtail

The initial Log Backend was based on a local file system. To accommodate distributed characteristics, we developed a highly reliable and low-latency Log Service as the new Log Backend. We abstracted a virtual backend to adapt to different log backends, connecting to various backends through the development of lightweight drivers.

Optimizing Log Management in Distributed Systems: Advanced Log Backend and Transaction Handling Mechanisms

The Write Ahead Log (WAL) is a technology related to the atomicity and durability of databases.It functions by converting random writes into sequential read-writes during transaction commits. Changes in transactions occur randomly across various pages, which are scattered.

What is WAL? and How to Apply it?(Commit Pipeline&Checkpoint)

A database is a "warehouse that organizes, stores, and manages data according to data structure." It is a collection of large amounts of data that are organized, shareable, and uniformly managed, and stored long-term in a computer.

What is the Binder in Database Kernels: A useful guide might helps you.

The previous sections primarily discussed MatrixOne's transaction processing. Previously, MO only supported the SI (Snapshot Isolation) isolation level, primarily implemented through MVCC (Multi-Version Concurrency Control), where data is multi-versioned. MatrixOne now supports the RC (Read Committed) isolation level for transactions.

Revolutionizing Data Integrity(2): Advanced Pessimistic Transactions in Distributed Databases

Previously, MatrixOne only supported optimistic transactions based on SI. It now supports pessimistic transactions and the RC (Read Committed) isolation level. Transactions with RC and SI can run concurrently in a single MatrixOne cluster.

Revolutionizing Data Integrity(1): Advanced Pessimistic Transactions in Distributed Databases

We are excited to announce MatrixOne 1.1.0 release! MatrixOne is a hyper-converged cloud-native database. It is designed to provide a cloud-native, high-performance, highly scalable, MySQL-compatible HTAP database. MatrixOne enables users to handle mixed workloads such as transactions, analytics, time-series, and streaming processing through a one-stop data processing solution.

Release Note: MatrixOne 1.1.0

This week's article might be a bit long. But we provide an in-depth look at MatrixOne's columnar storage layout, examining its design, structure, and data access methods since version 0.5. We focus on the rationale for its columnar approach, addressing the layout's design challenges, and exploring its adaptability for diverse workloads.

MatrixOne's Layout Mastery: Navigating Design, Structure, and Data Access

As we recap the journey of developing our query plan system, it's essential to highlight the pivotal transition from the foundational work on the Binder to the advanced stages involving the Planner and Optimizer. This shift marked a significant progression in both the complexity of tasks and the depth of technical innovation required.

Golang in Action (2):Mastering Planner and Optimizer in Query Plan Development

In versions prior to 0.4, MatrixOne's computation engine was implemented based on a factorization approach. However, this approach lacked versatility, such as the inability to support non-equi-join conditions. Therefore, starting with version 0.5, we officially decided to abandon the factorization approach and develop a new computation engine from scratch.

Golang in Action (1): How to Quickly Implement a Query Plan

This article would have a brief explanation of the read and write request process in logservice.

Advanced Operations in Log Service: Write&Read

The this article will cover advanced concepts such as the bootstrap process, the heartbeat mechanism, and the role of the Replicated State Machine (RSM) in the MatrixOne Log Service.

Advanced Operations in Log Service: Bootstrap, Heartbeat, and RSM

This article will provide an in-depth look at the overall architecture of the MatrixOne Log Service, detailing its design principles and the integration of the dragonboat library and raft protocol. We will also introduce on the client and server aspects of the MatrixOne Log Service. It will offer detailed explanations of client functionalities and server-side request processing.

Log Service Essentials: Architecture,WAL Management, Client and Server Operations

From the insights gained in the last article, which highlighted the challenges and limitations of MatrixOne's early architecture, we transition into this section, where the focus shifts to the transformative rebirth and strategic refinement of its structure.

How to Overcome Three Major Challenges (2): Transitioning from NewSQL to HTAP Distributed Architecture

In recent years, HTAP databases have become a fashionable term, and mentioning HTAP has turned into a trend among many database professionals. How to create an HTAP database, starting from the architectural level to address and embrace future changes, remains an ongoing exploration for many database companies.

How to Overcome Three Major Challenges (1): Transitioning from NewSQL to HTAP Distributed Architecture

After Optimizing Golang Performance (1) Memory Related and Golang Profiling, we are now facing the last topic of this tutorial.

Optimizing Golang Performance (3): How to write performance tests?

In the pervious blog, we exploration into Golang's memory-related features, but it is just the beginning. In forthcoming articles, we will delve deeper into various facets of Golang, uncovering more advanced topics and best practices.

Optimizing Golang Performance (2):Golang Profiling

Nowadays, with the booming cloud native ecosystem driven by Kubernetes, Docker, and many other star projects, more and more startups as well as big companies in China have started to pay attention to Golang.

Optimizing Golang Performance (1): Memory Related

MatrixOne is a cloud native database that not only supports high-performance analytical queries on ultra-large-scale datasets, but also has high-throughput, low-latency transactional read and write capabilities. This paper introduces the architecture of TAE (Transactional Analytical Engine), the storage engine of MatrixOne database.

TAE-MatrixOne Cloud Native Transaction and Analytics Engine

Golang provides a pprof performance profiling tool, allowing us to quickly and roughly analyze performance bottlenecks, which is widely used in daily development. However, as performance tuning requirements continue to rise, we need more accurate performance metrics, such as the execution time of a particular Golang function. At which point pprof can't help, and we have to find another way.

Dynamic Tracing Using BPF in Golang

Migrating to cloud-native and taking a distributed architecture/design approach to creating cloud-native applications has been a major trend in recent years, and this trend is accelerating further. There is no more important driver than the ability to dramatically reduce application downtime, high resiliency, high resource utilization, etc.

MatrixOne Chaos Testing Way

The database kernel is extensive, but the process of executing SQL is pretty much the same.

An In-Depth Look at MatrixOne Parser

In the era of digital transformation, data has become increasingly valuable. There is a growing trend of convergence in data technology. To address the pain points of customers, we have launched a simple, fast, and cloud native database, MatrixOne.

MatrixOne System Architecture

Our followers may find that the performance of MatrixOne 0.5 is much slower than before. In the release article of MatrixOne 0.2 earlier last year, there were reports that the single-table performance of MatrixOne SSB exceeded ClickHouse, and the multi-table performance reached StarRocks. So why did it become slower in version 0.5?

Why MatrixOne 0.5 Become Slower?

TAE (Transactional Analytical Engine) is the storage engine for MatrixOne, named so because it needs to support both TP and AP. The initial version of TAE implementation was released with MatrixOne 0.5, which was a single node storage engine. Starting from MatrixOne 0.6, TAE will transition into a cloud native distributed architecture with separate computing and storage.

Transactional Analytical Engine

We are thrilled to announce the release of MatrixOne 1.0.0-RC1! This milestone marks we have basically achieved the initial design goal of the MatrixOne open-source project. Cheers!

Release Note: MatrixOne 1.0.0-RC1

MatrixOne is a new generation of hyper-converged heterogeneous databases committed to creating a single architecture to handle mixed workloads such as OLTP, OLAP, streaming etc. MatrixOne is developed by Go language and has been open-sourced since October 2021. As a database implemented in Go competing with top-tier OLAP databases written in C++, MatrixOne has applied many optimizations including high-performance hash table implementation, which will be explained in detail in this article.

Go Language Advanced Practices: MatrixOne's Journey in Designing and Implementing High-Performance Hash Tables

The SQL execution engine of the database is responsible for processing and executing SQL requests. Generally, the query optimizer outputs a physical execution plan, which is usually composed of a series of operators. To ensure efficient execution, operators need to be composed into a pipeline.