Unlocking the Secrets of Data Anomalies: Tencent’s Groundbreaking Database Research

In recent work, Tencent Cloud’s TDSQL database team has made core advances in the fundamental research of data anomalies.

They conducted systematic research proving that data anomalies are infinite in number, and developed an open‑source mini‑program that can detect any type of data anomaly; the research has also been patented.

Why is this research important?

The core significance lies in uncovering the nature and intrinsic patterns of data anomalies, revealing the essence of isolation levels and concurrency algorithms, enabling systematic study and improvement of various concurrency control algorithms.

In simple terms, data anomalies serve as a key to unlocking concurrency control techniques.

Compared with traditional case‑by‑case studies, this new approach offers many advantages, as shown in the diagram below.

“Data anomaly” receives its first formal definition in theory.

In the database field, a data anomaly is defined as follows: in the history of concurrent transactions, if there exists a directed cycle based on conflict‑serializable techniques, it is called a data anomaly.

This definition allows a clear statement of transaction consistency: if no data anomaly exists, the transaction is considered consistent.

Data anomalies are classified into three new categories:

Write Anomaly (WAT): the cycle contains a write‑write precedence.

Read Anomaly (RAT): the cycle contains one or more write‑read precedences but no write‑write precedence.

Cross Anomaly (IAT): anomalies other than write or read anomalies.

Further detailed classifications include:

Single‑Data Anomaly (SDA): occurs on a single variable across two transactions.

Dual‑Data Anomaly (DDA): occurs on two variables across two transactions.

Multi‑Data Anomaly (MDA): all other anomalies beyond SDA and DDA.

To help engineers quickly learn about data anomalies, Tencent Cloud has released an open‑source mini‑program “Quick Learning of Data Anomalies” with features such as Linux support, precise anomaly detection, multiple detection algorithms, and a user‑friendly interactive interface.

Recent years have seen Tencent’s research focus on core database theory, including systematic studies of distributed consistency, transaction consistency integration, and the redefinition of transaction consistency.