Architecture Workshop(架构工作坊)
知识库位置:
FDE Knowledge Base/07 Playbooks/05-Architecture-Workshop.md
Playbook Metadata(手册元数据)
Section titled “Playbook Metadata(手册元数据)”Playbook ID: PB-005
Stage(所属阶段):Solution Design(方案设计)
Primary Goal(主要目标):Build shared understanding of target architecture and key decisions.(对目标架构与关键决策形成共同理解。)
Typical Duration(典型时长):90–120 minutes
Primary Audience(主要使用者):FDE + Solution Architect + Customer IT
Difficulty(难度):Advanced(高级)
Status: CoreVersion: v0.1Objective(目标)
Section titled “Objective(目标)”Architecture Workshop 的目标不是:
现场把每个组件都设计完。
而是:
- 对目标架构形成共同理解。
- 明确关键技术决策。
- 识别需要验证的风险。
- 确认 Prototype(原型)/ Pilot(试点)如何实现。
Inputs(输入)
Section titled “Inputs(输入)”进入 Workshop 前需要:
- Business Use Case(业务用例)
- Requirements(需求)
- Target Workflow(目标工作流)
- System Landscape(系统全景)
- Data Inventory(数据清单)
- Security Constraints(安全约束)
- Technical Risks(技术风险)
Participants(参与者)
Section titled “Participants(参与者)”- FDE
- Solution Architect(解决方案架构师)
- Engineering Lead(工程负责人)
- Customer IT(客户信息技术部门)
- Security(安全团队)
- Data Owner(数据责任人)
- Business Representative(业务代表)
Workshop Flow(工作坊流程)
Section titled “Workshop Flow(工作坊流程)”Step 1 — Restate the Problem(重述问题)
Section titled “Step 1 — Restate the Problem(重述问题)”第一张图不要是架构。
先写:
User(用户):Problem(问题):Business Outcome(业务结果):Constraints(约束):Step 2 — Define Boundaries(定义边界)
Section titled “Step 2 — Define Boundaries(定义边界)”明确:
- 哪些系统属于 Solution(解决方案)
- 哪些是 External Dependency(外部依赖)
- 哪些是 Existing Enterprise Systems(既有企业系统)
Step 3 — Select Architecture Pattern(选择架构模式)
Section titled “Step 3 — Select Architecture Pattern(选择架构模式)”参考:Reference Architecture Selection Guide(参考架构选型指南)
例如:
Knowledge Problem(知识问题)→ RAG(检索增强生成)
Known Process + AI(已知流程 + AI)→ Workflow Automation(工作流自动化)
Dynamic Actions(动态动作)→ Agent(智能体)Step 4 — Draw High-Level Architecture(画总体架构)
Section titled “Step 4 — Draw High-Level Architecture(画总体架构)”先控制在:
5–10 major components(5–10 个主要组件)避免一上来画 50 个框。
Step 5 — Data Flow(数据流)
Section titled “Step 5 — Data Flow(数据流)”明确:
Where does data enter?(数据从哪里进入?)Where is it transformed?(在哪里转换?)Where is it stored?(存储在哪里?)Who can see it?(谁能看到?)Where does it leave?(从哪里离开?)Step 6 — Identity & Security(身份与安全)
Section titled “Step 6 — Identity & Security(身份与安全)”必须明确:
User(用户)↓Identity(身份)↓Authorization(授权)↓Data / Tool(数据 / 工具)Step 7 — Failure Paths(失败路径)
Section titled “Step 7 — Failure Paths(失败路径)”逐个问:
- LLM(大语言模型)不可用?
- CRM 不可用?
- Search(搜索)不可用?
- Queue(队列)积压?
- Human Reviewer(人工复核员)不在?
Step 8 — Key Decisions(关键决策)
Section titled “Step 8 — Key Decisions(关键决策)”记录 ADR(Architecture Decision Record,架构决策记录)。
例如:
Use RAG instead of fine-tuning.(用 RAG 而非微调。)Use workflow instead of autonomous agent.(用工作流而非自主智能体。)Use batch instead of streaming.(用批量而非流式。)模板见 Architecture Decision Record Template(架构决策记录模板)。
Step 9 — Prototype Boundary(原型边界)
Section titled “Step 9 — Prototype Boundary(原型边界)”明确:
哪部分现在验证?
Prototype 不应该把整个 Production(生产)Architecture 全做完。
Outputs(产出)
Section titled “Outputs(产出)”- High-Level Architecture(总体架构)
- Data Flow(数据流)
- Integration Map(集成图)
- Security Model(安全模型)
- ADRs(架构决策记录)
- Technical Risks(技术风险)
- Prototype Scope(原型范围)
- Open Questions(待解问题)
Facilitation Principle(引导原则)
Section titled “Facilitation Principle(引导原则)”如果团队开始争论具体技术:
Kafka 还是某个 Queue(队列)?
先问:
这个组件需要满足什么业务和非功能要求?
避免工具先行。