The 5 Layers of a Modern Web Stack
A request's journey from the user's screen to the metal — and which tools live at each stop
· 웹 서비스 기술 스택 5계층 지도
① Presentation & Edge프레젠테이션 및 엣지
What the user sees, delivered fast — UI frameworks + CDN edge caching.
UI: React · Next.js · Flutter · Svelte | Edge/CDN: Cloudflare · Fastly · Vercel Edge
▼
② Integration & Messaging통합 및 메시징
How parts talk — synchronous APIs and asynchronous message brokers.
API: REST · gRPC · GraphQL | Async: Kafka · RabbitMQ · GCP Pub/Sub · SQS
▼
③ Business Logic & Data Access비즈니스 로직 및 데이터 액세스
The core rules of the app — server frameworks plus the efficiency layer in front of data.
Server: Spring · Django · Fastify/Express · Rails | Efficiency: Redis (cache) · Hibernate/Prisma (ORM)
▼
④ Storage & Analytics데이터 저장 및 분석
Where data lives and where it becomes insight — OLTP databases through big-data/ML tooling.
DB: PostgreSQL · MySQL · MongoDB | Analytics/ML: Spark · BigQuery · PyTorch
▼
⑤ Infrastructure인프라
The ground everything stands on — cloud hosting and container orchestration.
Cloud: AWS · GCP · Azure | Runtime: Docker · Kubernetes · Cloud Run
Where rabbit sits on this map rabbit 대입
① Next.js (App Router) — no CDN yet (Cloud Run serves directly) ·
② REST route handlers (`app/api/*`); no broker yet — the KB-RAG/MCP work (design §5) adds an MCP surface ·
③ Next.js server + `lib/*`; Prisma as the ORM ·
④ Cloud SQL Postgres (S6) ·
⑤ GCP Cloud Run + Docker (staging domain = design §11).
Gaps worth knowing: no CDN, no message broker, no analytics layer — all fine at rabbit's scale; the map shows what would be added first if scale demanded it.
How to use this map: when a new tool appears ("should I care about X?"),
first place it on a layer — most "X vs Y" debates (gRPC vs GraphQL, Kafka vs RabbitMQ) are
within-layer choices that never touch the other four layers.
rabbit · docs/knowledge · added 2026-07-17 (source: jay's 5-layer stack note pasted in session)