How Beike Guarantees High Availability in Complex Real‑Estate Transactions

Business Complexity and Technical Landscape

Beike’s real‑estate platform supports a wide range of services—second‑hand houses, new homes, renovation, leasing—requiring coordination among dozens of roles, multiple front‑end channels (Android, iOS, mini‑programs, PC) and back‑end systems. A single transaction can involve searching listings, brokerage, contract signing, mortgage, tax, registration, and post‑sale handover, creating long call chains, massive data volumes, and cross‑departmental dependencies.

The technical stack relies on micro‑services, gateways, Redis, message queues, Hadoop‑based big‑data components, and extensive use of open‑source libraries.

Key Quality Challenges

1. Micro‑service scale : More services and release cycles increase regression scope, environment instability, and the need for high performance and availability.

2. Historical technical debt : Diverse tech stacks and frequent refactoring amplify regression effort and quality loss.

3. Testing data and environment : Complex business scenarios require constructing multi‑entity test data (houses, customers, agents, contracts) and provisioning dozens of environments.

4. Cross‑domain quality solutions : Need to abstract common quality capabilities across mobile, web, and backend services.

Three‑Tier Quality Model

The quality philosophy separates visible quality (observable outcomes), hidden quality (processes such as CI pipelines, code reviews, performance testing), and invisible quality (design, architecture, requirement definition). This model guides the construction of a unified quality platform.

Core Quality Platforms

KeTest : A one‑stop quality solution providing automated regression, data management, environment provisioning, and performance testing for all business lines.

KeOnes : Connects development, CI/CD, and release stages, breaking down 28 production steps, integrating quality gates, and surfacing issues early.

sosotest : Supports multi‑protocol automated test case creation (HTTP, Dubbo, etc.), mock generation, and integrates with CI pipelines; currently hosts 46 000 test cases covering 87 % of business lines.

KeDiff : A traffic‑replay diff testing platform that captures logs, replays requests in baseline and comparison environments, filters noise, and generates diff reports; has processed over 2 600 regression runs and 13.7 million cases.

Virtual City : A sandbox environment that mirrors end‑to‑end business flows, providing 5 000+ test accounts for scenario verification, new‑business validation, and demo purposes.

Environment Governance : A Kubernetes‑based test‑container cloud offering unified configuration, one‑click environment cloning, and resource isolation; supports 1 000+ projects and 1 600+ active environments.

Data Construction Platform : Enables low‑threshold, flexible, and visual data generation via SQL, Dubbo, HTTP, Kafka, Redis, etc., allowing non‑technical QA to create complex test data quickly.

KePTS : Performance testing platform built on Grinder and Groovy, automating test data extraction from logs, providing templates, and scaling to full‑link load tests; improves testing efficiency by >30 % and has intercepted over 150 performance issues.

KeMTC : Mobile testing suite offering crash monitoring, automated case execution, cloud device usage analytics, and process‑level efficiency improvements.

Quality Governance and Culture

Beike has institutionalized quality metrics (Quality 1.0 whitepaper), regular examinations for engineers, a “fire‑fighting” incident response team, and a double‑loop quality improvement system that combines team‑level diagnostics with individual incentives (quality points, certifications).

Activities include large‑scale crowd‑testing, AI‑assisted image recognition for intelligent testing, chaos‑engineering drills, and continuous DevOps maturity assessments.

Future Directions

Plans focus on intelligent testing (AI‑driven image recognition, automated diff selection), expanded chaos‑engineering practices, and refined DevOps certification to further raise reliability and delivery speed.