📊 HiddenMerit Morning Post · Issue 83

Focus on Database Frontiers, Practical Insights for DBAs August 23, 2026 ｜ 5 Selected Global Breaking News

01｜Domestic Database Market Share Breaches 30% Tipping Point: Dameng, Kingbase, OceanBase Three‑Strong Pattern Established

According to the latest CCID Consulting and China Merchants Securities data, the domestic database software market reached RMB 7.93 billion in 2025, with a market share of 26.89%, and is expected to exceed RMB 10 billion in 2026, approaching a 30% market share. Dameng ranked first among domestic vendors with 13.2%, followed by Renda Kingbase at 11.5% and OceanBase at 10.8%. The combined share of these three in China’s relational database market has exceeded 35%. The three‑strong pattern of Dameng, Kingbase, and OceanBase is now essentially established, and core system replacement in key industries such as finance, telecommunications, and government will become the main battlefield in the next phase.

DBA Perspective: The breach of the 30% tipping point means domestic databases are shifting from “policy‑driven” to “market‑driven,” and DBA job demand in Xinchuang positions across key industries such as finance, telecommunications, and government will further expand. Dameng, Kingbase, and OceanBase should be the priority directions for Xinchuang skill development.

02｜DTCC 2026: Tencent Cloud Officially Releases TDSQL Nexa, Four AI Database Capabilities Enter Production

From August 20‑22, the 17th China Database Technology Conference (DTCC 2026) was held in Beijing. 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. Tencent Cloud Vice President Wang Yicheng stated that in the AI era, databases need four core capabilities: multi‑modal unified hosting, agent‑native architecture, semantics and memory, and governance and autonomy.

Core Capabilities and Production‑Grade Data:

TDSQL Nexa: Over 10x computational performance improvement and over 60% cost reduction compared to open‑source combination solutions. Nexa Knowledge semantic layer improves the accuracy of agent‑generated SQL by over 50% .
Agent Memory: Open‑sourced for 80 days with over 20,000 GitHub Stars; after integration, task success rates improved from 60% to 80% with a 66.3% cost reduction. The new “Team Memory Edition” supports cross‑agent knowledge sharing.
DatabaseClaw: Already in large‑scale production use at lead customers, with some fault root cause localisation completed in as little as 3 minutes.
TDSQL‑C: Already integrated into AI applications such as WorkBuddy, Yuanbao, and Hunyuan, stably serving 500,000 developers and processing billions of requests daily.
DBA Perspective: TDSQL Nexa’s semantic layer solves the problem of agents “not understanding business context.” DatabaseClaw’s 3‑minute root cause localisation provides DBAs with a quantitative benchmark for evaluating AI operations tools. Agent Memory’s 20,000 Stars data also indicates that AI agent memory storage is becoming a new core database workload.

Source: DTCC 2026 Official Information, IT168 Coverage

03｜GBASE GBase 8s Releases HK‑Tree Bitmap Index at DTCC: Designed for HTAP Mixed Workloads

At DTCC 2026, GBASE GBase 8s Product Line General Manager Yang Wei delivered a keynote on “A BITMAP Index for HTAP and Kernel Innovations Around Production Environments,” systematically introducing GBase 8s’ self‑developed HK‑Tree (High‑density Key Tree) index technology.

Technical Breakthrough: Traditional B+Tree has low storage efficiency and insufficient index‑only scans in low‑cardinality scenarios. Traditional bitmap indexes improve storage efficiency but cannot support high‑concurrency DML operations. HK‑Tree aims to “enable indexes with high‑speed writes, high‑speed reads, and high‑density storage simultaneously” – it can directly replace traditional B+Tree in all low‑cardinality scenarios, delivering exponential performance advantages in wide‑table multi‑column query scenarios.

Real‑World Test Data: Based on testing with 360 million rows, HK‑Tree demonstrated excellent performance in inserts, updates, wide‑table multi‑index queries, and aggregation queries across cardinality ranges of 100 to 5,000, covering equality queries, range queries, multi‑condition combined queries, and typical analytical operations including COUNT, DISTINCT, and GROUP BY.

TAC High Availability Solution: GBase 8s also released the TAC (Transaction Augmented Continuity) solution, achieving transparent application‑level database node switchover – users only need to configure any node IP in the URL without distinguishing between primary and standby. Through three layers of continuity guarantee – session continuity, transaction continuity, and current SQL connection continuity – business interruption during switchover is eliminated.

[quads id="805"]

DBA Perspective: HK‑Tree bitmap indexes provide DBAs with a new tuning tool for low‑cardinality wide‑table scenarios. In Xinchuang selection, HK‑Tree’s support for HTAP mixed workloads is worth particular attention – it allows a single system to handle high‑concurrency OLTP while efficiently supporting analytical queries. TAC’s transparent switchover directly addresses the “application error” concern that DBAs fear most during core system high‑availability failover.

Source: GBASE DTCC 2026 Release

04｜CVE-2026-55642 (CVSS 9.8): dbx‑web Database Client Default No Password, Public Exposure Equals Compromise

In August, the dbx‑web database client was found to have a critical security vulnerability (CVE-2026-55642), with a CVSS score of 9.8 (Critical). The vulnerability allows unauthenticated remote attackers to execute arbitrary SQL queries, affecting all versions up to v0.5.50.

Vulnerability Details: dbx‑web is the Web/Docker service component of the t8y2/dbx cross‑platform database client. The authentication middleware is “fail‑open” when password_hash is None – requests pass authentication checks unconditionally. More critically, dbx‑web binds to 0.0.0.0 (all network interfaces) by default, and the official quick‑start process does not require setting a default password. Attackers can access API endpoints such as /api/connection/connect and /api/query/execute to execute arbitrary SQL using stored database credentials, achieving full read, modification, and destruction capabilities.

Remediation: Upgrade to v0.5.51 or higher. Temporary mitigation: do not expose dbx‑web/Docker services directly to the public internet; restrict access to local or trusted internal networks only; explicitly configure a strong password; and audit existing deployments for anomalous SQL activity.

DBA Perspective: CVSS 9.8 means this vulnerability has an extremely low exploitation barrier and high impact. As a database client tool web version, if deployed by DBAs or developers for remote management, it could become a springboard for attackers to enter enterprise internal networks. Particularly concerning is the dual design flaw of “default no password + binding to all interfaces” – the documentation does not mandate setting a password, leaving many out‑of‑the‑box deployments default‑exposed. DBAs are advised to immediately check whether dbx‑web services are deployed internally, upgrade versions, and implement strict network access controls on publicly exposed management tools.

Source: VulDB, NIST NVD

05｜OceanBase Seeks Series A Funding: Targeting RMB 2‑3 Billion to Expand AI Database, Annualized Revenue Exceeds RMB 1.4 Billion

According to Bloomberg and Guandian reports, Ant Group’s OceanBase is in talks with investors for a Series A funding round of approximately RMB 2‑3 billion to enable more independent operations and expand AI database services. OceanBase’s annualized revenue this year exceeds RMB 1.4 billion (approximately $200 million), up 70% year‑on‑year, and ranked first in China’s distributed database market last year according to IDC.

OceanBase views San Francisco‑based Databricks and other companies as competitive targets. Databricks stated in February this year that its annualized revenue is on track to reach $5.4 billion, with AI products contributing $1.4 billion.

DBA Perspective: OceanBase’s first external funding is a landmark event for domestic databases moving toward “market‑oriented independent operations.” Benchmarking against Databricks means its strategic direction is extending from a “database company” to an “AI data platform” – DBA skills need to expand from “distributed database operations” to “AI data platform architecture and multi‑modal data management.”

Source: Bloomberg, Guandian

📚 SQL Little Knowledge Point

This Issue’s Knowledge Point: What are the “Four Core Capabilities of AI Databases”?

Tencent Cloud Vice President Wang Yicheng proposed at DTCC 2026 that databases in the AI era need the following four core capabilities:

Capability	Description	Corresponding Product
Multi‑Modal Unified Hosting	Simultaneous support for structured, semi‑structured, and unstructured data, covering transactions, vectors, full‑text, etc.	TDSQL Nexa
Agent‑Native Architecture	Permission, connection, and audit systems designed for agent access patterns	TDSQL Boundless, TDSQL‑C
Semantics and Memory	Enabling databases to understand business semantics and沉淀 long‑term memory	Nexa Knowledge, Agent Memory
Governance and Autonomy	Row/column‑level permissions, operation auditing, AI‑driven intelligent operations	DatabaseClaw

AI agents are becoming the new users of databases – in future selection, enterprises will simultaneously evaluate a database’s ability to support both core business systems and agent applications. For DBAs, this means mastering new skills in both “AI for DB” (using AI to optimise databases) and “DB for AI” (enabling databases to support AI applications).

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