architecture-research

Software Architecture Research

Overview

Architecture research is a systematic process for evaluating technology options, design patterns, and system architectures. Create Architecture Decision Records (ADRs) to document decisions with context, rationale, and trade-offs. Use this skill to make informed technology choices backed by real research.

When to Use This Skill

This skill should be used when:

• Evaluating technology stack options for a new project
• Researching design patterns and architectural approaches
• Comparing cloud services and infrastructure options
• Creating Architecture Decision Records (ADRs)
• Documenting technology trade-offs and rationale
• Researching integration patterns and best practices

Visual Enhancement with Project Diagrams

When documenting architecture decisions, always include diagrams.

Use the project-diagrams skill to generate:

• C4 model diagrams (Context, Container, Component)
• System architecture diagrams
• Integration pattern diagrams
• Deployment architecture diagrams
• Data flow diagrams

python .claude/skills/project-diagrams/scripts/generate_schematic.py "diagram description" -o diagrams/output.png

---

Architecture Research Process

Phase 1: Context Gathering

Before researching solutions, understand the problem space.

Requirements Analysis

Functional Requirements:

• What must the system do?
• What are the core use cases?
• What integrations are required?

Non-Functional Requirements (Quality Attributes):

Attribute	Questions to Answer
Performance	Response time targets? Throughput requirements?
Scalability	Expected growth? Peak loads? Scaling dimensions?
Availability	Uptime requirements? Recovery time objectives?
Security	Authentication needs? Data sensitivity? Compliance?
Maintainability	Team size? Skill levels? Update frequency?
Cost	Budget constraints? Operational cost limits?

Constraints Identification

Technical Constraints:

• Existing systems to integrate with
• Required technology compatibility
• Infrastructure limitations
• Data residency requirements

Organizational Constraints:

• Team skills and experience
• Timeline pressures
• Budget limitations
• Vendor preferences or restrictions

Phase 2: Technology Research

Research options systematically using real documentation and benchmarks.

Research Protocol

For each technology option:

1. Official Documentation Review

   • Use research-lookup to find official docs
   • Verify current version and roadmap
   • Check feature completeness for requirements

2. Production Case Studies

   • Search for "X in production" experiences
   • Look for similar-scale deployments
   • Find post-mortems and lessons learned

3. Benchmark Data

   • Search for performance benchmarks
   • Find comparison studies
   • Verify benchmark relevance to your use case

4. Community Health

   • Check GitHub stars, commits, contributors
   • Review issue resolution time
   • Assess documentation quality

5. Total Cost of Ownership

   • Licensing costs
   • Infrastructure requirements
   • Operational overhead
   • Training and hiring costs

Technology Comparison Matrix

Create a comparison matrix for each decision area:

comparison:
  category: "Database"
  options:
    - name: "PostgreSQL"
      pros:
        - "ACID compliance"
        - "Rich feature set"
        - "Strong community"
      cons:
        - "Vertical scaling primarily"
        - "Operational complexity at scale"
      fit_score: 8  # 1-10
      sources:
        - "Official PostgreSQL documentation"
        - "Benchmark: TPC-C results"
    - name: "MongoDB"
      pros:
        - "Horizontal scaling"
        - "Flexible schema"
        - "Developer friendly"
      cons:
        - "Eventual consistency trade-offs"
        - "Memory usage"
      fit_score: 6
      sources:
        - "MongoDB documentation"
        - "Case study: Company X migration"

Phase 3: Design Pattern Research

Research architectural patterns appropriate for the problem.

Common Architectural Patterns

Pattern	When to Use	Trade-offs
Monolithic	Small team, simple domain, rapid MVP	Scaling limits, deployment coupling
Microservices	Large team, complex domain, independent scaling	Operational complexity, network overhead
Event-Driven	Async workflows, decoupled systems, audit trails	Eventual consistency, debugging complexity
Serverless	Variable load, simple functions, cost optimization	Cold starts, vendor lock-in, state management
CQRS	Read/write asymmetry, complex queries	Complexity, eventual consistency
Hexagonal	Testability, multiple interfaces, long-lived systems	Initial overhead, abstraction complexity

Pattern Evaluation Checklist

For each pattern considered:

• Does it address the core requirements?
• Does team have experience with this pattern?
• What are the operational implications?
• What are the testing implications?
• What are the scaling characteristics?
• What are the failure modes?

Phase 4: Architecture Decision Records (ADRs)

Document decisions in ADR format for future reference.

ADR Template

# ADR-NNN: [Title]

## Status
[Proposed | Accepted | Deprecated | Superseded by ADR-XXX]

## Context
What is the issue that we're seeing that is motivating this decision or change?

## Decision Drivers
- [Driver 1]
- [Driver 2]
- [Driver 3]

## Considered Options
1. [Option 1]
2. [Option 2]
3. [Option 3]

## Decision
We will use [Option X] because [reasoning].

## Rationale
Detailed explanation of why this option was chosen over alternatives.

### Option 1: [Name]
**Pros:**
- Pro 1
- Pro 2

**Cons:**
- Con 1
- Con 2

**Evidence:**
- [Source/benchmark/case study]

### Option 2: [Name]
[Same structure]

## Consequences

### Positive
- [Consequence 1]
- [Consequence 2]

### Negative
- [Consequence 1]
- [Mitigation strategy]

### Risks
- [Risk 1] - Mitigation: [Strategy]

## Related Decisions
- [ADR-XXX: Related decision]

## References
- [Link to documentation]
- [Link to benchmark]
- [Link to case study]

ADR Best Practices

Do:

• Write ADRs as you make decisions, not after
• Include context that future readers will need
• Link to evidence and sources
• Document rejected options and why
• Keep ADRs immutable (supersede, don't edit)

Don't:

• Write ADRs retrospectively without context
• Omit trade-offs or negative consequences
• Make decisions without research evidence
• Leave ADRs in "Proposed" status indefinitely

Phase 5: C4 Model Documentation

Use the C4 model to document architecture at multiple levels.

Level 1: System Context Diagram

Shows the system in context with users and external systems.

Include:

• System under design (single box)
• User types/personas
• External systems
• Relationships and data flows

python .claude/skills/project-diagrams/scripts/generate_schematic.py \
  "C4 Context diagram: [System Name] in center. \
   Users: [User types]. \
   External systems: [Systems]. \
   Arrows showing relationships." \
  -o diagrams/c4_context.png

Level 2: Container Diagram

Shows high-level technology choices.

Include:

• Applications and services
• Databases and data stores
• Message queues
• Technology choices annotated

python .claude/skills/project-diagrams/scripts/generate_schematic.py \
  "C4 Container diagram: \
   Web App (React), API (Node.js), Database (PostgreSQL). \
   Show communication protocols between containers." \
  -o diagrams/c4_container.png

Level 3: Component Diagram

Shows components within a container.

Include:

• Major components/modules
• Their responsibilities
• Dependencies between components

Level 4: Code (Optional)

UML class diagrams or code structure for complex components.

Research Output Templates

Technology Research Report

# Technology Research: [Category]

## Executive Summary
[2-3 sentence summary of recommendation]

## Requirements Addressed
- [Requirement 1]
- [Requirement 2]

## Options Evaluated

### Option 1: [Name]
**Overview:** [Brief description]
**Fit Score:** [X/10]

**Strengths:**
- [Strength 1]
- [Strength 2]

**Weaknesses:**
- [Weakness 1]
- [Weakness 2]

**Evidence:**
- [Source 1]: [Finding]
- [Source 2]: [Finding]

**Cost Estimate:** [Monthly/Annual]

[Repeat for each option]

## Comparison Matrix

| Criterion | Weight | Option 1 | Option 2 | Option 3 |
|-----------|--------|----------|----------|----------|
| Performance | 25% | | | |
| Scalability | 20% | | | |
| Cost | 20% | | | |
| Team Fit | 15% | | | |
| Ecosystem | 10% | | | |
| Risk | 10% | | | |
| **Total** | 100% | | | |

## Recommendation
[Recommended option with justification]

## References
1. [Source 1]
2. [Source 2]

Architecture Overview Document

# Architecture Overview: [Project Name]

## System Context
[C4 Context diagram]
[Description of system boundaries and external dependencies]

## High-Level Architecture
[C4 Container diagram]
[Description of major components and technology choices]

## Key Architecture Decisions

| Decision | ADR | Status |
|----------|-----|--------|
| Database Selection | ADR-001 | Accepted |
| Authentication | ADR-002 | Accepted |
| Deployment Platform | ADR-003 | Proposed |

## Quality Attributes

### Performance
- Target: [Metrics]
- Approach: [Strategy]

### Scalability
- Target: [Metrics]
- Approach: [Strategy]

### Security
- Requirements: [List]
- Approach: [Strategy]

## Technology Stack

| Layer | Technology | Rationale |
|-------|------------|-----------|
| Frontend | | |
| API | | |
| Database | | |
| Cache | | |
| Queue | | |
| Infrastructure | | |

## Integration Points

| System | Integration Type | Protocol | Notes |
|--------|-----------------|----------|-------|
| | | | |

## References
- [Link to detailed ADRs]
- [Link to C4 diagrams]

Research Best Practices

Do's

• Use research-lookup for every major decision
• Cite real sources, not assumptions
• Include both pros and cons for all options
• Consider total cost of ownership, not just licensing
• Document decisions as ADRs immediately
• Generate diagrams for visual communication

Don'ts

• Don't make technology choices based on hype
• Don't ignore team skill gaps
• Don't assume "industry standard" without verification
• Don't skip documenting rejected options
• Don't make permanent decisions with temporary data

Final Checklist

Before completing architecture research:

• All major technology areas researched
• At least 3 options evaluated per decision
• Real sources cited for all recommendations
• ADRs created for all significant decisions
• C4 diagrams generated for system overview
• Trade-offs explicitly documented
• Costs estimated with sources
• Team skill gaps identified
• Risks documented with mitigations