2nd/10_Wiki/Topics/AI_and_ML/Software Architecture Pattern.md

id, title, category, status, canonical_id, aliases, duplicate_of, source_trust_level, confidence_score, verification_status, tags, raw_sources, last_reinforced, github_commit, tech_stack

id	title	category	status	canonical_id	aliases	duplicate_of	source_trust_level	confidence_score	verification_status	tags	raw_sources	last_reinforced	github_commit	tech_stack
wiki-2026-0508-software-architecture-pattern	Software Architecture Pattern	10_Wiki/Topics	verified	self	Architecture Patterns System Design Patterns	none	A	0.9	applied	architecture patterns system-design microservices serverless		2026-05-10	pending	language framework multi kubernetes/serverless/edge

Software Architecture Pattern

매 한 줄

"매 architecture pattern 의 problem-driven choice — layered (CRUD), microservices (org / scale), event-driven (decouple), hexagonal (testability), serverless / edge (ops scale)". 매 2026 의 reality 의 hybrid — modular monolith + selective microservices + event backbone (Kafka / NATS) + edge (Cloudflare Workers / Deno Deploy / Vercel). 매 single-pattern dogma X — 매 trade-off 의 explicit.

매 핵심

매 Major patterns

• Layered (n-tier): presentation / business / data — 매 CRUD default.
• Microservices: 매 service 의 independent deploy + own DB.
• Event-driven (EDA): 매 producer-consumer + queue / log.
• Hexagonal (ports & adapters): 매 domain 의 IO 의 isolate.
• CQRS + Event Sourcing: read / write split, append-only event.
• Modular Monolith: 매 monolith 위 의 boundary — 매 underrated 2026 default.
• Serverless: FaaS, 매 ops-zero, cold-start trade.
• Edge / Distributed: 매 user-near, 매 SQLite-replicated (Turso, D1).

매 Cross-cutting

• API Gateway: 매 frontend의 single entry.
• BFF: 매 client-specific aggregator.
• Saga: 매 distributed transaction.
• Outbox: 매 db + event 의 atomic publish.
• CRDT: 매 multi-region eventual consistency.

매 Modern (2026) trends

• Modular monolith 의 first — 매 microservices 의 premature 의 X.
• Edge runtime (Workers, Lambda@Edge) 의 read path.
• AI-augmented architect — Claude / GPT-5 의 ADR draft / pattern fit review.
• Event streaming default — Kafka / Redpanda / NATS JetStream.

매 응용

1. SaaS multi-tenant — modular monolith + tenant DB.
2. E-commerce — microservices + saga + event backbone.
3. Real-time collab — CRDT + edge + WebSocket.
4. AI app — gateway + queue + worker (LLM inference).

💻 패턴

Hexagonal (ports & adapters, Python)

from typing import Protocol

# 매 port (domain interface)
class OrderRepo(Protocol):
    def save(self, order: "Order") -> None: ...
    def by_id(self, id: str) -> "Order | None": ...

# 매 domain (pure)
class Order:
    def __init__(self, id, items): self.id, self.items = id, items
    def total(self): return sum(i.price for i in self.items)

class PlaceOrder:
    def __init__(self, repo: OrderRepo): self.repo = repo
    def __call__(self, order: Order):
        if order.total() <= 0: raise ValueError("empty")
        self.repo.save(order)

# 매 adapter (Postgres impl) — 매 domain 의 zero coupling
class PgOrderRepo:
    def __init__(self, conn): self.conn = conn
    def save(self, order): self.conn.execute("INSERT ...", ...)
    def by_id(self, id): ...

CQRS + Event Sourcing (skeleton)

from dataclasses import dataclass

@dataclass(frozen=True)
class OrderPlaced: order_id: str; total: float

class OrderAggregate:
    def __init__(self):
        self.id = None; self.total = 0
    def apply(self, event):
        if isinstance(event, OrderPlaced):
            self.id = event.order_id; self.total = event.total

class EventStore:
    def __init__(self): self.events = []
    def append(self, e): self.events.append(e)
    def load(self, id):
        agg = OrderAggregate()
        for e in self.events:
            if getattr(e, "order_id", None) == id: agg.apply(e)
        return agg

Outbox pattern (atomic db + event)

BEGIN;
INSERT INTO orders (id, total) VALUES ('o1', 100);
INSERT INTO outbox (aggregate_id, type, payload, created_at)
  VALUES ('o1', 'OrderPlaced', '{"id":"o1","total":100}', NOW());
COMMIT;

-- 매 worker 의 outbox 의 poll → Kafka publish → mark sent

Saga (orchestration, TypeScript)

async function placeOrderSaga(order: Order) {
  const compensations: (() => Promise<void>)[] = [];
  try {
    await reserveStock(order);
    compensations.push(() => releaseStock(order));
    await chargePayment(order);
    compensations.push(() => refund(order));
    await scheduleShipment(order);
  } catch (e) {
    for (const c of compensations.reverse()) await c().catch(() => {});
    throw e;
  }
}

Edge handler (Cloudflare Workers + D1)

export default {
  async fetch(req: Request, env: { DB: D1Database }) {
    const url = new URL(req.url);
    if (url.pathname === "/api/post" && req.method === "GET") {
      const id = url.searchParams.get("id")!;
      const row = await env.DB.prepare("SELECT * FROM posts WHERE id = ?")
        .bind(id).first();
      return Response.json(row, {
        headers: { "cache-control": "public, max-age=60, s-maxage=300" },
      });
    }
    return new Response("not found", { status: 404 });
  },
};

Modular Monolith (Go module structure)

cmd/api/main.go
internal/
  billing/        # 매 module — own domain, own table prefix
    api.go
    domain.go
    repo.go
  catalog/
    api.go
    ...
  shared/
    events.go     # 매 cross-module event bus

Event-driven (NATS JetStream)

import { connect, JSONCodec } from "nats";
const nc = await connect({ servers: "nats://nats:4222" });
const js = nc.jetstream();
const jc = JSONCodec();

// publish
await js.publish("orders.placed", jc.encode({ id: "o1", total: 100 }));

// subscribe (durable consumer)
const c = await js.consumers.get("ORDERS", "ship-worker");
const iter = await c.consume();
for await (const m of iter) {
  await ship(jc.decode(m.data));
  m.ack();
}

API Gateway (Kong declarative)

_format_version: "3.0"
services:
  - name: orders
    url: http://orders.internal:8080
    routes: [{ paths: ["/orders"], strip_path: false }]
    plugins:
      - name: rate-limiting
        config: { minute: 60, policy: redis, redis_host: redis }
      - name: jwt

매 결정 기준

상황	Pattern
Small team, CRUD	Layered or Modular Monolith
Org scaling, 매 indep deploy	Microservices (with care)
Audit / replay	Event Sourcing + CQRS
Read-heavy global	Edge + cache
Distributed transactions	Saga + Outbox
Real-time collab	CRDT + WebSocket
Spiky / unpredictable	Serverless

기본값: Modular Monolith first. 매 split 의 evidence-driven (org boundary + perf SLO).

🔗 Graph

• 부모: System-Design
• 변형: Microservices · Event-Driven-Architecture · Hexagonal Architecture · CQRS
• 응용: Modular Monolith
• Adjacent: API-Gateway · Service Mesh · Edge Computing

🤖 LLM 활용

언제: ADR draft, 매 pattern fit review (problem → pattern recommendation), trade-off matrix generate. 언제 X: actual capacity planning (load test), team-specific tech-stack lock-in (org context).

❌ 안티패턴

• Premature microservices: 매 distributed monolith — 매 worst.
• Distributed transaction (2PC): 매 saga 의 use.
• Shared DB across services: 매 microservices 의 violate.
• No outbox: 매 dual-write inconsistency.
• Pattern dogma: 매 trade-off 의 explicit X.

🧪 검증 / 중복

• Verified (Fowler, Vernon, Newman, Richardson architecture references).
• 신뢰도 A.

🕓 Changelog

날짜	변경
2026-05-08	Phase 1
2026-05-10	Manual cleanup — full pattern catalog with code examples