Cloud Architecture Design
Core concepts
Alibaba Cloud architecture questions test whether you can translate requirements into a defensible topology. Know the building blocks: Regions and Zones, VPC with vSwitches, ECS instance families (general, compute, memory, burstable), ESS auto scaling, SLB/ALB/NLB load balancers, RDS with read-only instances, and OSS for object storage. Understand the shared-responsibility split and how managed services (RDS, Redis, SLB) shift operational burden away from you.
Design patterns to recognize
High availability across multiple zones within one Region; horizontal scaling behind a load balancer; decoupling via Message Queue or Serverless; caching with ApsaraDB for Redis; and separating stateless compute from stateful data. Know when to choose a managed database over self-managed ECS, and when Serverless (Function Compute) beats provisioning servers.
Traps
Questions often hide a constraint that invalidates the obvious answer: a single-zone requirement, a cost ceiling, or a latency SLA that forces a multi-Region active-active design. Watch for "most cost-effective" versus "highest availability" - they lead to different answers. Also beware of assuming a managed service spans Regions automatically; most are Regional.
Rule of thumb
Design for failure at the smallest blast radius first: multi-zone before multi-Region, and only add cross-Region replication when the RPO/RTO explicitly demands it. If a requirement mentions surviving a full Region outage, multi-zone alone is never sufficient.