📊 HiddenMerit Morning Post · Issue 81

Focus on Database Frontiers, Practical Insights for DBAs August 21, 2026 ｜ 5 Selected Global Breaking News

01｜DTCC 2026 Day 2: Kingware Upholds RAC Path, Yashan Proposes New “Knowledge Engine” Paradigm

On August 21, the 17th China Database Technology Conference (DTCC 2026) entered its second day, with deep integration of AI and databases emerging as the central theme.

Kingware’s “Change and Constancy” Discourse: CETC Kingware Product Manager Zhang Hao delivered a keynote titled “Changing Scenarios, Unchanging Core: Continuous Evolution of Converged Database Architecture in the AI Era.” He noted that AI prosperity presents a two‑pronged challenge for databases – downward, core transaction systems demand extreme data consistency and business continuity; upward, scenario diversification brings challenges from multi‑modal data, semantic retrieval, and mixed workloads. Zhang Hao explicitly asserted that “AI can optimise database operations, diagnostics, and even SQL generation, but AI cannot solve the classic problem of ‘multiple nodes writing to the same data concurrently while guaranteeing ACID’ – this problem can only be solved by the database kernel itself.” Kingware’s KES RAC shared‑storage multi‑active architecture remains the objectively optimal solution for core systems.

Yashan’s “Knowledge Engine” New Direction: Yashan Technology Partner Wang Nan, speaking in the DB4AI track, argued that although large model capabilities have matured, “the data exists, but AI cannot read or use it.” The core viewpoint: knowledge modelling, representation, and retrieval are the true keys to AI application implementation. Yashan proposed a paradigm shift from “program + data” to “agent + knowledge.”

• DBA Perspective: Kingware’s steadfastness on RAC architecture provides important decision‑making reference for DBAs in choosing between distributed and shared‑storage cluster architectures. Yashan’s “knowledge engine” direction suggests that competition in AI data foundations will shift from “storage capability” to “knowledge modelling capability,” requiring DBAs to pay attention to emerging technology areas such as knowledge graphs and semantic retrieval.

02｜Tencent Cloud Launches TDSQL Nexa: Agent‑Native Data Foundation with Multiple AI Capabilities Entering Production

On the first day of DTCC 2026, Tencent Cloud officially released TDSQL Nexa, an AI‑era data foundation that connects dispersed multi‑source enterprise data through a unified entry point, providing agents with business semantic understanding and permission governance capabilities. Multiple AI database capabilities have already entered production environments in Tencent’s internal AI products and enterprise deployments.

Core Capabilities:

• Unified Data Platform: Integrates transaction processing, full‑text and vector retrieval, AI computing, and large‑scale analytics. Testing shows over 10x computational performance improvement and over 60% cost reduction compared to open‑source combination solutions.
• Nexa Knowledge Semantic Layer: Identifies and沉淀 business knowledge including table names, column names, and metric definitions, providing business context for agent queries. In medium‑scale application scenarios, agent‑generated SQL requires virtually no manual modification, with average business performance improvement of over 50% .
• Security Governance: Supports row/column‑level permissions and operation auditing, with high‑risk operations intercepted or routed to manual approval.

AI Database Capability Panorama: Tencent Cloud has formed a complete system covering agent data, retrieval, memory, and intelligent operations. Agent Memory, open‑sourced for 80 days, has already gained over 20,000 GitHub Stars, improving task success rates from 60% to 80% with a 66.3% cost reduction. DatabaseClaw has entered large‑scale production use at lead customers, with some fault root cause localisation completed in as little as 3 minutes.

• DBA Perspective: The release of TDSQL Nexa is further evidence of the trend toward “database as AI data platform.” The semantic layer addresses business semantic understanding challenges, providing DBAs with a reference for evolving from “data custodian” to “knowledge asset沉淀.” Agent Memory’s open‑source metrics (20,000 Stars, +20% success rate) provide DBAs with quantitative references for evaluating AI memory solutions.

03｜Dameng All‑in‑One Appears at DTCC: From Extreme Performance to Intelligent Computing, Xinjiang Mobile Customer Shares Real‑World Experience

At DTCC, Dameng All‑in‑One Product Line General Manager Zhu Honglei systematically introduced the latest achievements of the DAMENG PAI Database All‑in‑One. In performance dimensions, Dameng has completed three core technology breakthroughs – extreme storage (tiered storage + intelligent caching), compute offloading (embedding compute engines in storage), and full‑stack acceleration (200Gb/s RoCE v2 high‑speed network) – aiming to “build a performance benchmark for domestic data services” and drive critical industries from “usable” to “true replacement and true use.”

Three AI Intelligent Computing Dimensions:

• Intelligent Computing Foundation: DM9 multi‑modal database kernel, with three‑tier intelligent caching technology delivering all‑flash‑equivalent performance at near‑HDD cost.
• Open Platform: Provides MCP Server and Open API dual pathways, enabling resource discovery, tool invocation, and controlled SQL for AI agents.
• Intelligent Operations: Builds a 7×24 online DBA expert team to achieve autonomous operation of the intelligent computing foundation.

Customer Practice: Xinjiang Mobile will share its complete practice path at the conference – from technology selection to flexible migration, from traditional operations to AI‑empowered operations – explaining why it chose Dameng for its core system database replacement.

• DBA Perspective: Dameng all‑in‑one’s “three‑tier intelligent caching” and “compute offloading” technologies provide quantifiable performance references for DBAs in core system selection. Xinjiang Mobile’s real‑world sharing (from selection to migration full chain) is an important reference for DBAs evaluating domestic database implementation capability in operator core scenarios.

04｜Tencent Cloud AI Database Panorama: Four Core Capabilities – Multi‑Modal Unification, Agent‑Native, Semantics and Memory, Governance and Autonomy

At DTCC, Tencent Cloud systematically articulated its agent‑oriented database strategy. Tencent Cloud Vice President Wang Yicheng stated, “The changes brought by agents are not just a new type of database user – data forms, workloads, contextual requirements, and security boundaries are also changing.” In the AI era, databases need four core capabilities: multi‑modal unified hosting, agent‑native architecture, semantics and memory, and governance and autonomy.

Production‑Grade Validation Data:

• TDSQL‑C: Already integrated into AI applications such as WorkBuddy, Yuanbao, and Hunyuan, stably serving 500,000 developers and processing billions of requests daily.
• Agent Memory: Open‑sourced for 80 days with 20,000 GitHub Stars, improving task success rates from 60% to 80% with a 66.3% cost reduction.
• Team Memory Edition: New cross‑agent sharing of conversations, code, Skills, and project knowledge.

Wang Yicheng predicted: “When selecting databases in the future, enterprises will not just look at replacing a core system foundation, but will simultaneously evaluate its ability to support both core business systems and agent applications – this will become a new differentiator in database selection.”

• DBA Perspective: Tencent Cloud’s four‑capability framework provides DBAs with a structured reference for evaluating “AI‑ready” databases. Production‑grade data from 500,000 developers and billions of requests validates the large‑scale maturity of cloud‑native databases in AI workload scenarios. Agent Memory’s 20,000 Stars and +20% success rate data provide DBAs with quantitative references for AI agent memory storage solution selection.

05｜DTCC 2026 Conference Recap: AI + Database Becomes the Absolute Core Theme, 17+ Tracks Cover Full Chain of Data‑Intelligence Integration

From August 20‑22, the 17th China Database Technology Conference (DTCC 2026) was held in Beijing. The conference, hosted by IT168 in partnership with the ITPUB and ChinaUnix technical communities, was themed “Integrating Data, Aggregating Intelligence, Creating the Future,” featuring 2 main venues, 17+ specialised technical tracks, and 100 top industry experts sharing insights.

Core Content Themes:

• Data+AI Full‑Chain Integration: Three‑part series covering low‑cost enterprise large model implementation, AI‑native databases, vector retrieval, and AI agent data engineering.
• AI for DB Intelligent Autonomous Operations: Focused on operations agents, multi‑agent collaboration, and ontology‑driven fault localisation, showcasing AIOps implementation in education, telecommunications, internet, and other industries.
• Database Kernel and Xinchuang Autonomous Control: Real‑world domestic replacement practices in finance, telecommunications, and insurance core systems, plus all‑in‑one transformation practices.
• “DBA Night” Expert Dinner: Closed‑door discussion on “AI’s Impact on the Full IT Stack and Response Strategies.”

The speaker lineup covered leading internet companies, the six largest state‑owned banks, the three major telecom operators, central SOE manufacturing DBA teams, and academic scholars.

• DBA Perspective: DTCC is the largest technical conference in China’s database field and an important window for DBAs to access technology evolution directions. This year, “AI + Database” is the absolute core theme – with both AI for DB and DB for AI tracks running in parallel. The Data+AI Full‑Chain Integration and AI for DB Intelligent Operations tracks will directly impact the skill evolution path for DBAs over the next 3‑5 years.

📚 SQL Little Knowledge Point

This Issue’s Knowledge Point: What is an “AI‑First Database”?

DTCC 2026 featured, for the first time, a “Multi‑Agent Collaboration and the Ultimate Evolution of Databases” track, where “AI‑First Database” emerged as a focus of discussion. An “AI‑First Database” refers to a database architecture where AI capabilities are treated as a first‑principle design element – AI is not a functional layer “added onto” the database, but is integrated into the database kernel from the very beginning of design.

Core Characteristics of an AI‑First Database:

1. Native Vector and Semantic Capabilities: Vector retrieval is natively supported at the kernel level, not through plugins.
2. Agent‑Native Access Patterns: Permission models, connection management, and audit trails are redesigned for agent access patterns (high‑frequency small batches, cross‑session persistence).
3. Knowledge Modelling and Memory: Databases not only store data but also possess business knowledge modelling, representation, and retrieval capabilities (as exemplified by Yashan’s “knowledge engine” direction).
4. Autonomous Operations: AI‑driven automatic diagnosis, self‑healing, and optimisation are “built‑in capabilities” rather than external tools.

Implications for DBAs: The evolution toward AI‑First Databases means the DBA’s focus will gradually shift from “SQL optimisation” to “knowledge asset沉淀” and “AI agent policy management.” In future database selection, “AI‑readiness” will become a decision factor as important as “performance.”

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