58.com Database Architecture Design Overview

The article outlines 58.com’s database architecture design, focusing on availability, read/write performance, consistency, and scalability.

Availability design : Uses replication and redundancy, with master‑slave and dual‑master configurations to ensure high read and write availability, while addressing master‑slave inconsistency.

Read performance : Improves through extensive indexing (different indexes per database), adding read replicas, and employing a service‑cache‑data layer that avoids read/write separation and mitigates inconsistency.

Consistency solutions : Introduces middleware routing writes to the master and reads to the master within a synchronization window, or simply routes all operations to the master; also uses a double‑eviction cache strategy to resolve cache‑database mismatches.

Scalability : Describes rapid horizontal scaling by converting N databases to 2N using dual‑master “master‑as‑slave” mode, and discusses schema evolution via log‑based or dual‑write methods, as well as sharding strategies for single‑key, one‑to‑many, many‑to‑many, and multi‑key tables.

SQL practices : Advises limited use of complex queries, joins, subqueries, triggers, UDFs, and transactions due to performance impact, and presents techniques for pagination, IN queries, and non‑partition key queries in a sharded environment.

The article also revisits Codd’s 12 rules for relational databases and offers practical advice such as minimizing reliance on database‑specific features and defining entity relationships with clear ownership and cardinality.