基于单元的架构

Architecture

基于单元的架构图，展示地理路由到独立、自包含的单元（US-East、EU-West），每个单元拥有自己的网关、计算、数据库、缓存和消息队列，加上用于跨单元复制和全局配置的共享服务。该模板模拟超大规模系统使用的单元架构模式以实现爆炸半径隔离，一个单元的故障不会影响另一个单元的用户。对于设计需要极高可用性和故障隔离的系统的架构师至关重要。

下载 .fz 文件

完整 FlowZap 代码

GlobalRouter { # Global Router
n1: circle label:"User Request"
n2: rectangle label:"Geographic Routing"
n3: diamond label:"User Cell Assignment?"
n4: rectangle label:"Route to Cell A"
n5: rectangle label:"Route to Cell B"
n1.handle(right) -> n2.handle(left)
n2.handle(right) -> n3.handle(left)
n3.handle(right) -> n4.handle(left) [label="US-East"]
n3.handle(bottom) -> n5.handle(top) [label="EU-West"]
n4.handle(bottom) -> CellA.n6.handle(top) [label="Forward"]
n5.handle(bottom) -> CellB.n12.handle(top) [label="Forward"]
}
CellA { # Cell A (US-East)
n6: rectangle label:"Cell Gateway"
n7: rectangle label:"Compute Layer"
n8: rectangle label:"Cell Database"
n9: rectangle label:"Cell Cache"
n10: rectangle label:"Cell Message Queue"
n11: circle label:"Cell Response"
n6.handle(right) -> n7.handle(left) [label="Process"]
n7.handle(right) -> n8.handle(left) [label="Read/Write"]
n7.handle(bottom) -> n9.handle(top) [label="Cache"]
n7.handle(bottom) -> n10.handle(top) [label="Async"]
n8.handle(right) -> n11.handle(left)
}
CellB { # Cell B (EU-West)
n12: rectangle label:"Cell Gateway"
n13: rectangle label:"Compute Layer"
n14: rectangle label:"Cell Database"
n15: rectangle label:"Cell Cache"
n16: rectangle label:"Cell Message Queue"
n17: circle label:"Cell Response"
n12.handle(right) -> n13.handle(left) [label="Process"]
n13.handle(right) -> n14.handle(left) [label="Read/Write"]
n13.handle(bottom) -> n15.handle(top) [label="Cache"]
n13.handle(bottom) -> n16.handle(top) [label="Async"]
n14.handle(right) -> n17.handle(left)
}
SharedServices { # Shared Services (Cross-Cell)
n18: rectangle label:"Global Config Service"
n19: rectangle label:"Cross-Cell Replication"
n20: rectangle label:"Monitoring and Alerting"
n18.handle(right) -> n19.handle(left) [label="Sync Config"]
n19.handle(top) -> CellA.n8.handle(bottom) [label="Replicate"]
n19.handle(top) -> CellB.n14.handle(bottom) [label="Replicate"]
n19.handle(right) -> n20.handle(left) [label="Health"]
}

为什么需要这个工作流？

Traditional multi-region deployments share databases and services across regions, meaning a failure in one component can affect all users globally. Cell-based architecture creates fully independent, self-contained cells where each cell serves a subset of users—limiting the blast radius of any failure to only the users in that cell.

工作原理

Step 1: A global router assigns users to cells based on geographic location or account ID.
Step 2: Each cell contains its own gateway, compute, database, cache, and message queue.
Step 3: Cells operate independently—a failure in Cell A does not affect Cell B.
Step 4: Shared services handle cross-cell concerns like global configuration and monitoring.
Step 5: Cross-cell replication synchronizes data for disaster recovery.
Step 6: New cells can be added to handle growth without affecting existing cells.

替代方案

Multi-region active-active shares resources and has larger blast radius. Single-region with failover has longer recovery times. This template shows the cell architecture used by hyperscale systems like AWS and Azure.

Key Facts

Template Name	基于单元的架构
Category	Architecture
Steps	6 workflow steps
Format	FlowZap Code (.fz file)