GLOSSARY
Clear definitions of distributed database, cloud-native, and TiDB-specific terminology — written for engineers, not marketers.
Hybrid Transactional and Analytical Processing
A database architecture that handles both transactional (OLTP) and analytical (OLAP) workloads on the same system simultaneously. HTAP eliminates the need for separate databases and ETL pipelines, enabling real-time analytics on live transactional data. TiDB is purpose-built for HTAP workloads — learn how TiDB's HTAP architecture works in practice.
Online Transactional Processing
A class of database workloads characterized by high-frequency, short-duration transactions — inserts, updates, and deletes. OLTP systems prioritize low latency and high concurrency. Examples include e-commerce order processing and banking transactions.
Online Analytical Processing
A class of database workloads involving complex queries over large datasets for business intelligence and reporting. OLAP queries typically scan many rows and perform aggregations. Traditionally handled by separate data warehouses, but TiDB's columnar storage engine (TiFlash) brings OLAP capabilities to the same cluster — read how an HTAP database handles both at the same time.
Modern relational databases that deliver SQL and ACID transactions with improved horizontal scalability and availability compared to traditional RDBMS. “NewSQL” is outcome-focused and not inherently cloud-native by design, even though many deployments run in cloud environments.
Distributed SQL Database
A cloud-native SQL database that operates as one logical system while automatically partitioning, replicating, and routing data and queries across nodes for scale and fault tolerance. TiDB is a distributed SQL database, providing MySQL compatibility with transparent distribution and resilience. See why distributed SQL databases elevate modern application development — covering architecture, partitioning, and ACID at scale.
Ti Key-Value Store
The distributed transactional key-value storage engine that powers TiDB. TiKV stores data across multiple nodes using the Raft consensus algorithm for strong consistency. For a deep dive into TiKV's internals, see the TiKV overview in TiDB Docs. It is also available as a standalone CNCF graduated project for teams that need a distributed key-value store without the SQL layer.
TiFlash Columnar Storage
TiDB's columnar storage extension that enables real-time OLAP queries. TiFlash maintains a columnar replica of TiKV data asynchronously, allowing analytical queries to run on column-oriented storage for significantly better performance — without impacting transactional workloads on TiKV. See the TiFlash overview for deployment and configuration details.
Placement Driver
The metadata management component of a TiDB cluster. PD is a central component of the TiDB architecture, responsible for storing cluster topology, allocating transaction IDs (timestamps), and orchestrating data placement and load balancing across TiKV nodes. It uses etcd internally for distributed consensus.
Raft Consensus Algorithm
A distributed consensus algorithm used by TiKV to ensure data consistency across replicas. Raft elects a leader node for each data region, and all writes must be acknowledged by a majority of replicas before committing — providing strong consistency guarantees even in the event of node failures. For an in-depth look at how Raft is implemented inside TiKV, see Building a Large-scale Distributed Storage System Based on Raft.
Horizontal Scaling (Scale-Out)
Adding more nodes to a distributed system to increase capacity, as opposed to vertical scaling (adding more resources to a single node). TiDB scales horizontally — you add TiKV nodes to increase storage and throughput, and TiDB nodes to increase query concurrency, without downtime.
An architecture where a single database cluster serves multiple tenants (customers or teams) with isolated resources and data. TiDB Cloud supports multi-tenancy through resource groups and access controls, enabling shared infrastructure with tenant-level isolation.
The ability to automatically scale database resources up or down in response to workload changes. TiDB Cloud provides elastic scaling — clusters expand to handle traffic spikes and contract during off-peak hours, reducing cost without manual intervention.
Atomicity, Consistency, Isolation, Durability
The four properties that guarantee reliable database transactions. TiDB provides full ACID compliance at the distributed level — a transaction either commits across all nodes or rolls back entirely, even in multi-region deployments. See how ACID scales in a distributed SQL database — and why it matters beyond single-node MySQL.
TiDB Cloud brings HTAP, elastic scaling, and MySQL compatibility to a fully managed service.