PolarDB:PolarDB for MySQL on ENS

Architectural overview

PolarDB for MySQL on ENS features an architecture optimized for edge environments. It uses an independent storage architecture based on all-flash disks. Each node (one primary node and multiple read-only nodes) has its own dedicated storage space. This design ensures the portability of the database service. It also inherits core capabilities from the public cloud version of PolarDB, such as one primary with multiple read-only nodes and read/write splitting, to ensure high performance and high availability. The service architecture is illustrated in the following figure:

• One primary, multiple read-only nodes

  PolarDB utilizes a distributed cluster architecture. A cluster contains one primary node and up to seven read-only nodes (it is recommended to maintain at least one read-only node to ensure high availability). The primary node handles both read and write requests, while the read-only nodes handle only read requests. By default, PolarDB clusters provide a transparent, highly available, and adaptive load balancing capability at no extra cost. Using a single cluster endpoint, SQL requests are automatically routed to the appropriate nodes in the cluster, delivering aggregated, high-throughput concurrent SQL processing.

• Compute-storage decoupled architecture

  PolarDB is built on a design philosophy that separates compute and storage, meeting the critical need for elastic scaling in public cloud computing environments. The database compute nodes (Database Engine Servers) only store metadata, while data files, Redo Logs, and other data are stored on remote storage nodes (Database Storage Servers). Compute nodes only need to synchronize Redo Log metadata, which significantly reduces replication latency between the primary and read-only nodes. In the event of a primary node failure, a read-only node can be quickly promoted to become the new primary.

Core features

• Advanced MySQL compatibility and features

  • Compatibility: 100% compatible with native MySQL and ApasaraDB RDS for MySQL. You can migrate your MySQL database to a PolarDB for MySQL cluster without changing any application code or configurations.

  • High-value features: Includes single-writer/multi-reader, stable operation with tables containing over 6 billion rows, sub-second DDL, and flashback query.

• Orca (Redis protocol compatible)

  • Built on the PolarDB for MySQL architecture, Orca implements a one-primary-multiple-read-only-nodes architecture through physical replication. It also supports both horizontal and vertical scaling.

  • A single system supports both Online Transactional Processing (OLTP) and persistent key-value (KV) storage, eliminating the need for separate purchases.

Why choose PolarDB for MySQL on ENS

• Reduced latency with edge proximity

  Edge node networks are physically closer to end users. This placement avoids long-distance data transmission, which significantly reduces database access latency and improves the user experience. In addition, distributed data caching effectively relieves pressure on the central node.

• Meet Compliance with wild coverage

  For business needs in countries or regions not yet covered by the public cloud, or where data residency laws require data to remain within national borders, PolarDB for MySQL on ENS is an ideal solution. It is also ideal for meeting local legal requirements, such as data residency.

  Note

  Alibaba Cloud ENS is built on more than 3,200 CDN carrier edge nodes, covering more than 70 key countries and regions. It also provides more than 900 public cloud edge nodes worldwide.

• High performance

  Achieve high-performance read/write processing and efficient primary-replica synchronization through a deeply optimized database kernel and physical replication technology based on Redo Logs.

• High availability and reliability

  • Based on physical replication, our Semisynchronous replication (Semi-sync) uses efficient Redo Log synchronization between the primary and replica nodes. This process significantly improves the efficiency of semi-synchronous replication, thereby minimizing performance impact on the primary database. Compared to asynchronous replication, the performance loss under high-concurrency workloads is only about 10%.

  • Provide comprehensive security for data access, storage, and management through measures like IP whitelists, VPC networking, and multi-replica data storage.

• High elasticity

  You can add, remove, or change node specifications in minutes. You can also scale out storage online. The entire process is completed with no service interruptions.

• Lock-free backup

  Utilizing snapshot technology on the underlying storage layer, you can back up terabytes of data within minutes. This process does not require table locking and has minimal impact on your business operations.

How to get started

Create an ENS network and an edge cluster in the region where you want to deploy your cluster. See Quick start for edge clusters.