majiayu000 Architecture Patterns Knowledge
Pattern Selection Framework
When recommending a pattern, evaluate against these criteria:
Team Fit
- Team size < 5 → Strongly favor monolith or modular monolith
- Team size 5-15 → Modular monolith or bounded microservices
- Team size 15+ → Microservices viable if org supports it
- No DevOps/Platform team → Avoid Kubernetes-heavy architectures
- Junior-heavy team → Favor simpler patterns with clear boundaries
Operational Reality
- 1 DevOps engineer → Cannot sustain >3 independently deployed services
- No dedicated DBA → Avoid polyglot persistence
- Limited monitoring experience → Start with monolith + good observability
- Budget constrained → Serverless for spiky loads, monolith for steady
Requirement Signals
- Independent scaling needs → Microservices or serverless
- Strong consistency requirements → Monolith or synchronous microservices
- Event-heavy domain → Event-driven architecture
- Read/write asymmetry → CQRS
- Rapid experimentation needed → Modular monolith (fastest to change)
- Regulatory isolation → Service per compliance boundary
Pattern Comparison Quick Reference
Monolith
- Best for: Small teams, rapid development, strong consistency needs
- Risks: Scaling bottleneck, deployment coupling, team scaling limits
- Migrate to: Modular monolith → microservices (progressive)
Modular Monolith
- Best for: Teams 3-10, need clear boundaries without operational overhead
- Risks: Module coupling creep, single deployment unit
- Migrate to: Extract modules to services as needed
Microservices
- Best for: Large teams, independent scaling, polyglot needs
- Risks: Distributed system complexity, data consistency, operational overhead
- Prerequisite: CI/CD maturity, monitoring, team autonomy
Event-Driven
- Best for: Async workflows, decoupled services, audit trails
- Risks: Eventual consistency complexity, debugging difficulty, event schema evolution
- Prerequisite: Team comfort with async patterns, good monitoring
Serverless-First
- Best for: Spiky workloads, cost optimization at low scale, rapid prototyping
- Risks: Cold starts, vendor lock-in, debugging difficulty, cost at scale
- Prerequisite: Cloud-native comfort, stateless design skills
CQRS
- Best for: Read-heavy systems, different read/write models, event sourcing
- Risks: Complexity, eventual consistency, increased codebase size
- Prerequisite: Strong domain modeling skills
Red Flags by Pattern
- Microservices with shared database → "You have a distributed monolith"
- Event sourcing without replay testing → "You'll lose data"
- Serverless with persistent connections → "Wrong pattern for this workload"
- CQRS for simple CRUD → "Over-engineered"
- Monolith with 50+ developers → "Deployment bottleneck incoming"