Cloud Service Solution Design and Architecture
Domain 2: Cloud Service Solution Design and Architecture
This domain tests architecture judgement: turning availability, recovery, security, performance, and cost requirements into a Huawei Cloud design. Know when to use multi-AZ active-active, when to use cross-region DR, and how ELB, AS, RDS, DRS, SDRS, CBR, and OBS replication fit together.
Core rules
- Define RTO and RPO first. Multi-AZ synchronous replication addresses RPO near zero within a region; cross-region asynchronous replication has a different RPO and protects against region failure.
- Stateless tiers scale horizontally behind ELB and Auto Scaling; stateful tiers need replication, failover, or backup.
- Use CBR for backup and restore, SDRS for VM-level DR, DRS for database replication, and OBS cross-region replication for object durability.
- Security belongs in the design: VPC segmentation, WAF for HTTP(S), Anti-DDoS, KMS/DEW for keys, and CTS for audit.
Worked example
A payment API needs 99.95% availability, RPO 0 within the region, and RTO 30 minutes for a region loss. Design: ELB across two AZs, Auto Scaling with ECS in both AZs, RDS for MySQL primary/standby across AZs for synchronous RPO 0, and DRS asynchronous replication to a standby region for regional DR. CBR protects against logical corruption; SDRS is optional if whole-VM recovery is required. The 30-minute RTO drives runbook automation and DNS switchover, not just data replication.
Traps
- Claiming multi-AZ meets cross-region DR; a region outage takes both AZs.
- Treating backup as DR without testing restore time; RTO includes detection, decision, and DNS TTL.
- Ignoring quota and shared dependencies such as IAM, DNS, and image services in the DR region.