What Makes TiDB a Cloud‑Native HTAP Database? An In‑Depth Overview

TiDB Overview

TiDB is an open‑source distributed relational database designed by PingCAP that supports Hybrid Transactional and Analytical Processing (HTAP). It provides horizontal scalability, financial‑grade high availability, real‑time HTAP, cloud‑native deployment, and full compatibility with the MySQL 5.7 protocol and ecosystem, delivering a one‑stop OLTP, OLAP, and HTAP solution for large‑scale, high‑availability applications.

OLTP and OLAP Data processing is divided into two major categories: Online Transaction Processing (OLTP) for routine transactional workloads such as banking, emphasizing memory efficiency, command rate, and concurrency; and Online Analytical Processing (OLAP) for complex analytical queries, emphasizing data analysis, SQL execution, disk I/O, and partitioning.

Key Features of TiDB

Highly compatible with MySQL – most applications can migrate to TiDB without code changes.

Horizontal elastic scaling – add nodes to increase throughput or storage on demand.

Distributed ACID transactions – full support for standard ACID guarantees.

Financial‑grade high availability – Raft‑based majority election provides 100% strong consistency and automatic failover.

One‑stop HTAP solution – TiDB together with TiSpark delivers simultaneous OLTP and OLAP processing without complex ETL.

Cloud‑native SQL database – designed for public, private, and hybrid clouds, simplifying deployment and maintenance.

Vibrant community – active open‑source contributions and ecosystem.

Architecture Overview

TiDB follows the Spanner model and consists of three components: TiDB Server, Placement Driver (PD), and TiKV Server.

TiDB Server

The TiDB Server receives SQL requests, performs query logic, and uses PD to locate the appropriate TiKV nodes. It is stateless, does not store data, and can be horizontally scaled behind load balancers such as LVS, HAProxy, or F5.

PD Server

PD (Placement Driver) manages cluster metadata, performs scheduling and load balancing for TiKV, and allocates globally unique, monotonically increasing transaction IDs. A PD cluster should have an odd number of nodes, typically at least three in production.

TiKV Server

TiKV is a distributed transactional key‑value storage engine. Data is stored in regions, each covering a key range. Regions are replicated using the Raft protocol, forming Raft groups that ensure consistency and fault tolerance. PD schedules region placement across TiKV nodes.

TiSpark

TiSpark enables complex OLAP workloads by running Spark SQL directly on TiDB’s storage layer, leveraging TiKV’s distributed architecture and integrating with the broader big‑data ecosystem.

Typical Use Cases

Financial services requiring strong consistency, high availability, and disaster recovery across multiple data centers.

High‑concurrency OLTP workloads with massive data volumes, benefiting from TiDB’s compute‑storage separation and horizontal scaling.

Real‑time HTAP scenarios where simultaneous transaction processing and analytical queries are needed without costly ETL pipelines.

Data aggregation and secondary processing, allowing businesses to generate near‑real‑time reports directly via SQL instead of complex Hadoop pipelines.

Deployment Example

TiDB clusters can be deployed using TiUP, PingCAP’s unified deployment tool. Documentation: https://pingcap.com/docs-cn/stable/production-deployment-using-tiup/