How PolarDB Redefines Cloud‑Native Relational Databases

Relational Databases

Relational databases, born half a century ago, remain the backbone of modern IT, supporting most commercial applications alongside CPUs and operating systems.

From E.F. Codd’s 1970 relational model paper to the commercial releases of DB2, Oracle, and SQL‑Server, relational databases have consistently driven data management innovation.

Despite the emergence of many new database types, relational databases dominate because of the SQL standard, which offers a high‑level, non‑procedural interface that bridges computer science and human‑readable data management.

SQL Language

SQL, proposed in 1974 by Boyce and Chamberlin, provides a natural‑language‑like syntax for defining and manipulating data, abstracting storage details and enabling widespread adoption across business logic.

SQL has become the de‑facto standard for relational databases, with no comparable replacement in sight.

OLTP

Jim Gray’s 1976 work formalized transaction processing and consistency, leading to the OLTP model for everyday transactional workloads such as banking.

OLTP systems must satisfy ACID properties (Atomicity, Consistency, Isolation, Durability) and are measured by response time and throughput.

Open‑Source Database Ecosystem

While commercial databases like Oracle, SQL‑Server, and DB2 still lead the market, open‑source systems such as MySQL and PostgreSQL have grown rapidly, now surpassing commercial offerings in popularity.

Current Stage of Cloud Computing

Cloud 1.0 connected users to compute resources; Cloud 2.0 aims to evolve computing power within the cloud, improving efficiency and integration.

Alibaba Cloud’s PolarDB was created to address the limitations of traditional relational databases in shared‑cloud environments, following a similar evolution path as Amazon Aurora.

Why Alibaba Cloud Developed PolarDB

Cloud 1.0 solved cost and elasticity but left gaps in performance, scalability, and reliability for relational workloads. PolarDB adopts a spiral‑up evolution philosophy, delivering higher performance at comparable cost.

PolarDB Design Philosophy

PolarDB focuses on cloud‑hosted relational data while delivering lower cost, high OLTP performance, business continuity, elastic scaling, and data security through continuous technical innovation.

Alibaba Cloud RDS Product Architecture Evolution

RDS offers multiple MySQL‑based product lines:

Basic edition – compute‑storage separation, suitable for low‑cost, low‑availability scenarios.

High‑availability edition – Active‑Standby architecture with 99.95% SLA, read‑write separation, and shared‑nothing design.

Financial edition – Raft‑based strong consistency for finance‑grade reliability.

PolarDB Architecture

PolarDB employs a distributed cluster with storage‑compute separation, active‑active failover, and shared storage accessed via high‑speed RDMA networking.

Database nodes are MySQL‑compatible; the shared storage hosts data files and redo logs, routed through a user‑space file system to remote chunk servers.

Chunk servers replicate data using Parallel‑Raft to ensure consistency.

Key Technical Points

Shared Disk Architecture

Logical data resides on shared chunk storage, allowing all DB servers to access the same data blocks concurrently.

Physical Replication

By storing redo logs on shared storage, PolarDB eliminates the double‑fsync overhead of traditional MySQL binlog replication, enabling fast read‑only node scaling and sub‑30‑second failover.

RDMA over High‑Speed Networks

RDMA provides zero‑copy, low‑latency data transfer between compute and storage nodes, dramatically reducing CPU overhead and performance jitter.

Snapshot‑Based Physical Backup

Snapshots use copy‑on‑write block device techniques, allowing point‑in‑time recovery faster than traditional full‑plus‑binlog restores.

Parallel‑Raft Algorithm

Parallel‑Raft extends Raft by permitting parallel log commits, improving concurrency while preserving consistency for chunk server I/O.

Docker Containerization

DB compute nodes run in Docker containers, offering lightweight isolation and efficient resource utilization.

User‑Space File System

A user‑space file system implements POSIX‑like semantics without kernel‑level overhead, simplifying distributed storage integration.

In summary, PolarDB integrates advances in compute virtualization, high‑speed networking, shared‑block storage, distributed file systems, and physical replication to achieve a quantum leap in cloud‑native relational database performance.