📊 HiddenMerit Daily · Issue 45

Focus on Database Frontiers, Practical Insights for DBAs June 17, 2026 ｜ 5 Selected Global Breaking News

01｜OceanBase Launches AI Integrated Solution at Financial Expo: Nearly 70% of Trillion‑Yuan Banks Choose It, Ranked #1 in Financial Distributed Market for Three Consecutive Years

On June 16, the 2026 China International Financial Expo opened in Shanghai. Ant Group’s OceanBase launched its “Lakehouse‑Integrated AI Database for Financial Scenarios” and a financial AI integrated data solution, processing transactions, analytics, and multi‑modal data within a single engine.

Key Data:

• Market Position: According to IDC data, OceanBase has been ranked #1 in market share in China’s financial industry distributed database on‑premises deployment market for three consecutive years.
• Bank Coverage: Nearly 70% of banks with trillion‑yuan assets have deployed their core systems on OceanBase.
• Financial Institution Coverage: Has served over 400 financial institutions, covering state‑owned major banks, joint‑stock banks, city commercial banks, rural commercial banks, and others; 75% of leading insurance institutions, 80% of leading securities firms, and 60% of leading fund institutions have chosen OceanBase for their core business systems.

Core Capabilities of the AI Solution:

• Lakehouse Integration: A single engine uniformly processes transactions, analytics, and multi‑modal data, supporting hybrid search across vectors, text, and more.
• Data Branching: Built‑in data branching capability enables safe AI exploration and iteration in isolated environments.
• Internal Validation: Already implemented in internal Ant Group innovation projects such as Alipay AI Pay, Ant Lingguang, and Ant Afu.

OceanBase CEO Yang Bing said in his keynote at the Financial Expo main forum: “Over the past 15 years, our core work has been to handle the most critical business workloads in the financial industry.” He identified two forces reshaping the data foundation – the data consumer has changed (Gartner predicts that by 2028, one‑third of enterprise software interactions will be completed by agents), and the shape of data has changed (over 80% of global data is unstructured).

Bank of Beijing, as a representative of trillion‑yuan city commercial banks, shared its practice: over three years, it has completed the migration of more than 200 systems, covering core business, general business, and comprehensive office domains, and has沉淀 a replicable and scalable distributed database migration methodology.

• DBA Perspective: The choice of nearly 70% of trillion‑yuan banks means that large‑scale replacement by domestic distributed databases in financial core scenarios has moved from “pilot projects” to “batch delivery.” Bank of Beijing’s migration of over 200 systems in three years is among the fastest in the financial industry, providing DBAs with a reference cadence and scale benchmark for bank Xinchuang migration. The AI solution has already been implemented in internal projects such as Alipay AI Pay, confirming that AI databases are no longer just concepts but a production‑validated technology path. DBAs should focus on mixed‑load tuning under the “lakehouse‑integrated” architecture and security isolation strategies for data branching capabilities.

• CTO Perspective: OceanBase’s three consecutive years as #1 in the financial distributed market, combined with Bank of Beijing’s full‑business‑domain migration validation, confirms that the reliability of domestic distributed databases in financial core systems has been fully validated by leading institutions. Gartner’s prediction that one‑third of enterprise software interactions will be completed by agents by 2028 means financial CTOs must proactively build AI‑native capabilities into their data architecture planning.

• Investor Perspective: The choice of nearly 70% of trillion‑yuan banks and three consecutive years of market leadership provide strong performance endorsement for OceanBase’s valuation in the capital market. The launch of the financial AI integrated solution marks that OceanBase’s valuation logic is upgrading from “distributed database replacement” to “AI data infrastructure.”

02｜Dameng Hosts Institutional Investor Research: Native Vector in AI Database + All‑in‑One Performance Multiplied, Global Expansion Accelerates

On June 12, Dameng (688692) hosted research visits from Bosera Fund and Changjiang Securities, disclosing the latest progress in its AI database, next‑generation all‑in‑one machine, and overseas business.

AI Database Progress:

• Native Vector Data Type: Dameng’s database management system now natively supports vector data types, providing complete vector data functionality – creation, querying, indexing, import/export, etc.
• Dameng Qizhi AI Data Platform: Based on the AI4DB concept, it enables intelligent database operations, SQL optimisation, and intelligent query and index tuning, solving the low efficiency, high cost, and experience‑dependence of manual parameter tuning.
• Multi‑Modal Database: Supports unified management and collaborative processing of multiple data models – graph, relational, document, vector – and has developed Dameng MCP services.
• Future Judgement: “AI capability integration will become a standard feature of databases, driving the industry toward secure, trustworthy, efficient, intelligent, and multi‑modal unified evolution.”

Next‑Generation All‑in‑One Machine:

• Through compute offloading technology, data movement overhead is effectively reduced, with performance improving by multiples in some scenarios, significantly improving query response speed and overall processing capability under mixed workloads.

Global Expansion Layout:

• In July 2025, established a wholly‑owned subsidiary, Hainan Dameng International Data Technology Co., Ltd., in Hainan, leveraging the Hainan Free Trade Port as an international business headquarters.

• Held a new product launch in Hong Kong in May 2026, and will continue to leverage international exhibitions and industry exchanges to convert overseas pilot projects into large‑scale sales.

• DBA Perspective: Dameng’s native vector data type and support for unified management of graph, relational, document, and vector data mean DBAs can manage multi‑modal data within a single Dameng database, eliminating data movement between multiple database systems. The “multiples performance improvement” data for the new all‑in‑one machine provides a quantitative basis for DBAs’ performance arguments in hardware selection. The development of Dameng MCP services is also worth attention – MCP is a standard protocol for AI agents to connect to databases, and its maturity will directly affect the efficiency of data access for AI applications.

• CTO Perspective: Dameng’s clear judgement that “AI capability integration will become a standard feature of databases” provides CTOs with directional guidance for database selection. The global expansion has been initiated (Hainan subsidiary + Hong Kong launch), indicating that Dameng is moving from “domestic replacement” to “international competition,” providing a reference for assessing the long‑term development space of domestic databases.

• Investor Perspective: Dameng’s simultaneous progress in AI databases, next‑generation all‑in‑one machines, and global expansion validates the company’s布局 across three growth tracks. The combination of native vector types, multi‑model databases, and AI4DB gives Dameng a complete product matrix in the AI database track.

03｜OceanBase’s Liu Depeng: Building a Multi‑Modal Converged Unified Data Foundation is Key to Government and Enterprise AI Implementation

On June 16, at the “2026 People’s Data Conference” hosted by People.cn’s People Data, Liu Depeng, General Manager of Government and Enterprise Solutions at OceanBase Haiyang Database, delivered a keynote speech, proposing that government and enterprise digital transformation in the AI era requires building a multi‑modal converged unified data foundation.

Core Judgements:

• Data Becomes a Factor of Production: Data is transforming from a record‑keeping asset to a useful factor of production – no longer just static historical accumulation, but a key resource capable of actively creating business value.
• Data is the Key Bottleneck for AI Implementation: There are currently four obstacles between data and AI – fragmented data, AI’s lack of unified and usable data objects, AI’s higher requirements for data real‑time performance, and more prominent security and compliance issues.
• The First Priority of AI is Not Complex Algorithms: “If the underlying data is fragmented and chaotic, the return on AI investment may be inefficient, or even negative.”

OceanBase’s Response: Proposed the “lakehouse‑integrated” concept, building a data foundation for the AI era that supports unified data consumption and efficient use. For government and enterprise scenarios, it recently released an AI integrated solution for digital government, focusing on supporting scenarios such as “unified urban management,” “unified online services,” and “unified cross‑department collaboration.”

The Era of Human‑AI Collaboration Has Arrived: Collaborative work methods are shifting from hierarchical to flat dynamic collaboration; decision‑making paradigms are shifting from human experience‑based judgment to humans defining rules and AI assisting in decision execution; data is evolving from record‑keeping assets to factors of production.

• DBA Perspective: Liu Depeng’s judgement provides DBAs with a framework for understanding industry change. “Data is the key bottleneck for AI implementation” means that the DBA’s value positioning in AI projects is upgrading from “data maintainer” to “AI data foundation architect.” The “lakehouse‑integrated” unified data foundation requires DBAs to have the ability to manage structured, semi‑structured, and unstructured data uniformly.

• CTO Perspective: The “four data obstacles” identified by Liu Depeng – fragmented data, lack of unified data objects, insufficient real‑time performance, and security/compliance issues – are the core pain points for government and enterprise AI implementation. When planning AI data architectures, CTOs should prioritise database platforms with multi‑modal convergence capabilities rather than maintaining high‑complexity architectures that “patch together multiple databases.”

• Investor Perspective: OceanBase’s “lakehouse‑integrated” concept and the release of its digital government AI solution mark OceanBase’s strategic extension from “financial core replacement” to “government and enterprise AI data foundation.” The authoritative endorsement platform of the People’s Data Conference also enhances OceanBase’s brand recognition in the government and enterprise market.

04｜SelectDB Releases Agent Native Data Infrastructure Capability Panorama, Litefuse Officially Open‑Sourced

Recently, SelectDB held its AI product launch, focusing on directions such as ultra‑fast analytics, unified multi‑modal data management, Agent Native, Agent Observability, and Serverless elastic architecture. It systematically demonstrated the latest product evolution for AI and agent scenarios, attracting over 30,000 developers to watch online.

Core View: Ma Ruyue, CEO of Feilun Technology, stated: “Only by solving the four challenges of ultra‑fast, unified, Agent Native, and Cloud elasticity can we become the best analytics engine for the Agent era.”

Five Capability Dimensions:

1. Ultra‑Fast: Sub‑second response has evolved from optimisation to baseline. In agent scenarios, a single question may trigger dozens of queries, amplifying query latency by 10‑50x. SelectDB maintains a leading position in authoritative benchmarks such as ClickBench and TPC‑H.
2. Unified: Enterprises do not lack vector databases; they lack a unified AI data infrastructure. SelectDB has strengthened multi‑modal management and hybrid search capabilities to uniformly manage structured and unstructured data.
3. Agent Native: Released MCP Server and semantic layer capabilities – the semantic layer defines business, MCP connects all agents. With its built‑in MCP Server, mainstream agents such as Claude Code, Codex, and Cursor can directly access SelectDB.
4. Agent Observability: Officially released Litefuse, an observability platform for agent scenarios, and announced its open‑sourcing. Compared to Langfuse, Litefuse achieves up to 88% storage space savings and 5‑10x text retrieval performance improvement. StepFun has built an Agent Trace platform based on SelectDB, achieving a data closed loop from runtime observability to effectiveness evaluation.
5. Cloud Elasticity: Alibaba Cloud SelectDB Serverless Edition was officially commercialised in March 2026, supporting second‑level elastic scaling, shifting resource costs from fixed provisioning to pay‑as‑you‑go.

• DBA Perspective: SelectDB’s Agent Native positioning provides another perspective – analytical databases are evolving from “data sources for BI tools” to “data infrastructure directly consumed by agents.” The release of MCP Server means DBAs need to pay attention to the MCP protocol – the standard protocol layer between AI agents and databases – whose maturity will determine whether agents can access data efficiently. The open‑sourcing of Litefuse provides DBAs with a free tool for agent observability, useful for auditing AI agents’ database access behaviour.

• CTO Perspective: SelectDB’s “Ultra‑Fast, Unified, Agent Native, Cloud Elasticity” four‑dimension framework provides CTOs with a systematic framework for evaluating analytical databases in the AI era. The combination of MCP Server + semantic layer is an important path to solving the “business semantic understanding” challenge of Text‑to‑SQL. The open‑sourcing of Litefuse lowers the barrier to agent observability and is worth evaluating for introduction into AI application architectures.

• Investor Perspective: SelectDB, as the commercialisation company of Apache Doris, is fully布局 in the Agent Native direction, representing the evolution direction of open‑source analytical databases in the AI era. The open‑source strategy of Litefuse will expand its ecosystem influence, and customer practices such as StepFun validate the product’s suitability for AI scenarios.

05｜Weekly Security Vulnerabilities Focus: CVE-2026-22015 (MySQL Privilege Escalation), Delphix Multi‑Database Connector Privilege Escalation

Multiple database‑related security vulnerabilities were disclosed this week:

CVE-2026-22015 (Oracle MySQL Server Privilege Escalation) : Affects MySQL Server versions 8.0.45 and below, 8.4.8 and below, and 9.6.0 and below. The vulnerability is located in the Information Schema component and can be remotely exploited to achieve privilege escalation. A CVE has been assigned; it is recommended to upgrade to the fixed version immediately.

CVE-2026-8654 (Delphix Multi‑Database Connector Privilege Escalation) : Affects multiple connectors in Delphix Continuous Data, including IBM Db2 Connector, MongoDB Connector, PostgreSQL Connector, MySQL Connector, Oracle EBS Connector, SAP HANA Connector, CockroachDB Connector, Couchbase Connector, Cassandra Connector, YugabyteDB Connector, MSSQL on Linux Connector, and Oracle Backup Ingestion Connector. The vulnerability allows remote attackers to achieve privilege escalation. Upgrading the affected components is recommended.

• DBA Perspective: CVE-2026-22015 affects MySQL 8.0.45 and below, 8.4.8 and below, and 9.6.0 and below, covering most current MySQL production deployments. MySQL 8.0 is already EOL; unupgraded users should urgently assess the security risks of the 8.0 branch. CVE-2026-8654 is notable because it affects more than ten database connectors – as a data management platform, vulnerabilities in Delphix’s connectors mean that attackers could gain a cross‑database privilege escalation channel through data management tools such as data replication and backup. DBAs should identify whether Delphix components are in use and evaluate patch upgrade priorities.

• CTO Perspective: CVE-2026-8654 covers 12 database connectors, reflecting the expanding security risks at the “data management tool layer.” Tools like Delphix have high‑privilege cross‑database access capabilities; once compromised, the damage radius is extremely large. CTOs should establish security audit mechanisms for the “data management tool layer” and include connector‑class components in vulnerability management scope.

• Investor Perspective: Delphix connector vulnerabilities covering 12 databases indicate that the security risks of data management platforms are becoming a new attack surface. Enterprise customer demand for security auditing of data management tools will continue to grow, benefiting service providers offering security scanning and compliance auditing for data management platforms.

📚 SQL Little Knowledge Point

This Issue’s Knowledge Point: What is MCP (Model Context Protocol)?

MCP (Model Context Protocol) is a standard protocol open‑sourced by Anthropic at the end of 2025, designed to enable AI agents to securely and efficiently connect to various data sources and tools.

Core Positioning: MCP serves as a “standard interface layer” between AI agents and external data systems – just as USB‑C allows different devices to connect to a computer with a single cable, MCP allows different AI agents to connect to databases, APIs, file systems, and other data sources using a unified protocol.

Relationship with Databases:

• Traditional approach: Each AI agent needs to write custom connection code for each data source.
• MCP approach: Data sources provide an MCP Server, and AI agents access them uniformly through the MCP protocol.

SelectDB’s MCP Practice: With SelectDB’s built‑in MCP Server, mainstream agents such as Claude Code, Codex, and Cursor can directly access SelectDB without custom development for each agent.

Significance for DBAs:

• MCP will become the mainstream protocol for AI agents to access databases; DBAs need to understand MCP security configuration and permission models.
• The deployment and management of MCP Servers may become a new responsibility for DBAs in the AI era.

HiddenMerit Team Production Slogan: 绩优隐于内，金石启新程 | Hidden deep. Merit bold. Forge ahead.