Why Do Your Apps Crash? Alibaba’s High‑Availability Architecture Playbook

Why Does Your Application Crash?

In everyday life, app crashes are usually visible to users, but the underlying cause often lies in the invisible, complex server‑side (cloud) infrastructure.

Extremely Complex Backend

A mature cloud architecture may involve nearly 200 Alibaba Cloud products across compute, security, and enterprise services. Key nodes from client to server include CDN, dynamic acceleration, DDoS protection, WAF, 4/7 layer load balancers, service groups, caches, databases, middleware, and infrastructure layers. Each node has multiple configurable options, and failure at any point can render the service unavailable.

Lack of Pre‑Planning

Without prior capacity planning and readiness measures such as elastic scaling, protection, or circuit‑breaker mechanisms, sudden traffic spikes can cause instability; rushed scaling may exacerbate problems rather than solve them.

Alibaba Engineers’ High‑Availability Architecture Notes

Architecture Design

Visualization is essential. Using AHAS architecture awareness, engineers can map cloud resources, containers, and applications, revealing dependencies across servers, containers, and processes. This enables CMDB visualization, asset management, and multi‑dimensional views for migration, refactoring, and resource optimization.

Strong and weak dependency governance allows marking non‑critical services as weak dependencies, enabling graceful degradation and resource savings when the system reaches its throughput limits.

Capacity Planning

External network simulation testing (PTS) can quickly generate traffic models matching production scale, compatible with JMeter scripts or using a zero‑code visual composer. Full‑link testing routes production traffic to shadow storage for accurate capacity measurement while keeping test data isolated.

Business Monitoring

ARMS provides end‑to‑end monitoring across pages, databases, application performance, infrastructure resources, and business metrics, reducing troubleshooting time and cross‑team communication costs.

Online Traffic Control

AHAS probes enable traffic throttling, peak‑shaving, and graceful degradation without code changes, offering per‑resource protection rules that take effect instantly.

Daily Inspection

Advisor intelligently inspects cloud resources, identifies risks, and provides recommendations based on Alibaba Cloud’s best‑practice knowledge and SRE experience.

Regular Chaos‑Engineering Drills

AHAS’s fault‑injection module follows chaos‑engineering principles, allowing users to design multi‑dimensional failure scenarios across resources, services, containers, and the cloud platform, leveraging a rich library of fault cases to improve architecture, business continuity, and recovery capabilities.

Tool Overview

AHAS – Application High‑Availability Service for automatic architecture detection, fault injection testing, and one‑click traffic control.

PTS – Cloud‑based performance testing platform supporting API debugging, traffic simulation, and integrated monitoring.

Advisor – Intelligent advisor offering diagnostics and optimization suggestions for cloud resources, architecture, performance, and security.

Enterprise‑grade high‑availability solution – Proven by Alibaba’s Double‑11 traffic peaks, providing cost control, emergency response, and disaster‑avoidance capabilities.

ChaosBlade – Chaos‑engineering tool for injecting failures and improving system resilience.

Sentinel – Lightweight traffic‑control framework for flow limiting, circuit breaking, and overload protection.