DeanKelly751 Cloud Architecture
Purpose
Provider-agnostic decision frameworks for designing resilient, cost-effective cloud systems.
Well-Architected Pillars
| Pillar | Key question | Failure mode |
|---|---|---|
| Reliability | Does it recover from failures? | Single points of failure, no redundancy |
| Performance | Does it meet latency/throughput targets? | Under-provisioned, wrong service type |
| Security | Is the attack surface minimized? | Over-permissioned, public exposure |
| Cost | Are you paying for what you use? | Over-provisioned, idle resources |
| Operational excellence | Can you operate it effectively? | No automation, manual processes |
Architecture Patterns
| Pattern | Use when | Key benefit |
|---|---|---|
| Multi-AZ | Production workloads | Survives datacenter failure |
| Multi-region | Global users, regulatory, DR | Low latency everywhere, regional failure tolerance |
| Serverless | Variable/spiky traffic, event-driven | Pay per use, no capacity planning |
| Container orchestration | Consistent deployment, team autonomy | Portability, resource efficiency |
| Managed services | Team lacks operational expertise | Reduced ops burden |
Resilience Patterns
- Circuit breaker: Stop calling failing downstream, give it time to recover
- Bulkhead: Isolate failures to prevent cascade
- Retry with backoff: Handle transient failures
- Health checks: Detect and route around unhealthy instances
- Graceful degradation: Serve partial functionality when dependencies fail
Cost Optimization Heuristics
- Right-size before scaling out
- Use spot/preemptible for fault-tolerant workloads
- Reserved capacity for steady-state baseline
- Auto-scale for variable demand
- Delete idle resources (automated detection)
- Data transfer costs often dominate at scale
Anti-Patterns
- Lift and shift without redesign: Running on-prem patterns in cloud (misses cloud-native benefits)
- Distributed monolith: Microservices with synchronous dependencies on every request
- Ignoring data gravity: Compute far from data = latency + transfer costs
- Single-AZ production: One datacenter failure takes down the system
Related Skills
infrastructure/knowledge-twelve-factor— app design for cloudinfrastructure/execution-containerization— container-based deploymentobservability/knowledge-observability-pillars— observing cloud systemsreviewers/review-production-readiness— cloud deployment readiness