Databases 12 min read

How TDSQL Achieves Multi‑Level Strong Consistency with 4×‑3× Performance Gains

This article explains how Tencent's TDSQL database tackles the combined challenges of transaction and distributed consistency by introducing a multi‑level strong consistency model that delivers several‑fold performance improvements over Spanner, CockroachDB, and native Greenplum while preserving ACID guarantees.

ITPUB
ITPUB
ITPUB
How TDSQL Achieves Multi‑Level Strong Consistency with 4×‑3× Performance Gains

Problem Background

In distributed transaction processing a system must satisfy the consistency requirement of ACID (the C) and the consistency aspect of the CAP theorem while maintaining high transaction throughput. The combination creates a complex technical challenge because both transaction‑level anomalies (dirty read, non‑repeatable read, phantom, write‑skew, etc.) and distributed‑consistency anomalies (steal‑read, immortal‑write, causal‑reverse) can appear.

Academic Context

Research identifies “strict serializability” (also called external consistency) as a model that simultaneously satisfies transaction consistency and CAP consistency. However, there are dozens of isolation levels and roughly sixty distributed‑consistency models, making the design space large.

Industry Implementations

Google Spanner is the only widely deployed system that implements strict serializability, but its transaction throughput is relatively low for high‑performance workloads.

TDSQL Multi‑Level Consistency

TDSQL extends the theory by modeling consistency from concepts, formal models, operation semantics, transaction frameworks, data structures and algorithms. It defines and proves multiple strong‑consistency levels:

SER‑L (Linear‑serializable) : linear consistency combined with serializability.

SER‑S (Sequential‑serializable) : sequential consistency combined with serializability.

These levels eliminate anomalies such as write‑skew, steal‑read and guarantee serializable execution.

Performance Evaluation

Experiments on Google Cloud compare TDSQL with Spanner, CockroachDB and native Greenplum under comparable resources.

Against Spanner, TDSQL’s multi‑level consistency achieves >4× higher throughput.

Against CockroachDB, throughput is >2× higher.

Against native Greenplum, throughput is >3× higher.

With network latencies of 1.5 ms and 5 ms, TDSQL’s higher‑consistency isolation levels still outperform Greenplum’s read‑committed isolation.

Significance

TDSQL is the first engineering system that implements multiple strong‑consistency levels, demonstrating that the theoretical models are practical and can be realized with high performance, making strong consistency viable for real‑world distributed databases.

Future Work and Open Source

The multi‑level consistency stack, named 3TS , is being open‑sourced. The GitHub repository ( https://github.com/Tencent/TDSQL/tree/master/3TS) already provides implementations of 13 concurrency‑control algorithms together with documentation.

Original Source

Signed-in readers can open the original source through BestHub's protected redirect.

Sign in to view source
Republication Notice

This article has been distilled and summarized from source material, then republished for learning and reference. If you believe it infringes your rights, please contactadmin@besthub.devand we will review it promptly.

Distributed TransactionsTDSQLperformance benchmarkingstrong consistencyDatabase Researchmulti-level consistency
ITPUB
Written by

ITPUB

Official ITPUB account sharing technical insights, community news, and exciting events.

0 followers
Reader feedback

How this landed with the community

Sign in to like

Rate this article

Was this worth your time?

Sign in to rate
Discussion

0 Comments

Thoughtful readers leave field notes, pushback, and hard-won operational detail here.