---
id: messaging-nats-rabbitmq-comparison
title: NATS / RabbitMQ / SQS — 큐 비교
category: Coding
status: draft
source_trust_level: B
verification_status: conceptual
created_at: 2026-05-09
updated_at: 2026-05-09
tags: [messaging, nats, rabbitmq, sqs, vibe-coding]
tech_stack: { language: "TS / NATS / AMQP", applicable_to: ["Backend"] }
applied_in: []
aliases: [NATS JetStream, RabbitMQ, AMQP, SQS, message broker]
---

# NATS / RabbitMQ / SQS

> Kafka 만이 답이 아니다. **NATS = 가볍고 빠름, RabbitMQ = AMQP routing, SQS = AWS managed**. 처리량 / ordering / 영속성 / 운영 부담 4축으로 선택.

## 📖 핵심 개념
- NATS Core: pub/sub fire-and-forget.
- NATS JetStream: 영속, replay, exactly-once.
- RabbitMQ: AMQP, exchange + routing key + queue.
- SQS: AWS managed FIFO 또는 standard.

## 💻 코드 패턴

### NATS JetStream
```ts
import { connect } from 'nats';

const nc = await connect({ servers: 'nats://nats:4222' });
const js = nc.jetstream();
const jsm = await nc.jetstreamManager();

// Stream 생성
await jsm.streams.add({ name: 'ORDERS', subjects: ['orders.*'], storage: 'file', max_age: 7 * 24 * 3600 * 1e9 });

// Publish
await js.publish('orders.created', JSON.stringify({ id, userId }));

// Consumer
const sub = await js.consumers.get('ORDERS', 'order-projector');
const iter = await sub.consume();
for await (const m of iter) {
  await handle(JSON.parse(m.string()));
  m.ack();
}
```

### RabbitMQ (AMQP)
```ts
import amqp from 'amqplib';

const conn = await amqp.connect('amqp://rabbit:5672');
const ch = await conn.createChannel();

await ch.assertExchange('orders', 'topic', { durable: true });
await ch.assertQueue('orders.created.projector', { durable: true });
await ch.bindQueue('orders.created.projector', 'orders', 'orders.created');

// Publish
ch.publish('orders', 'orders.created', Buffer.from(JSON.stringify(order)), {
  persistent: true,
  messageId: eventId,
});

// Consume
await ch.prefetch(10);
ch.consume('orders.created.projector', async (msg) => {
  if (!msg) return;
  try {
    await handle(JSON.parse(msg.content.toString()));
    ch.ack(msg);
  } catch {
    ch.nack(msg, false, false); // DLX 로
  }
});
```

### RabbitMQ DLX (dead-letter exchange)
```ts
await ch.assertQueue('orders.created.projector', {
  durable: true,
  arguments: {
    'x-dead-letter-exchange': 'orders.dlx',
    'x-message-ttl': 60_000,
    'x-max-retries': 5,
  },
});
```

### AWS SQS (standard)
```ts
import { SQSClient, SendMessageCommand, ReceiveMessageCommand, DeleteMessageCommand } from '@aws-sdk/client-sqs';

const sqs = new SQSClient({});

await sqs.send(new SendMessageCommand({
  QueueUrl: queueUrl,
  MessageBody: JSON.stringify(order),
  MessageAttributes: { eventId: { DataType: 'String', StringValue: id } },
}));

// Long-poll
const r = await sqs.send(new ReceiveMessageCommand({
  QueueUrl: queueUrl,
  WaitTimeSeconds: 20,
  MaxNumberOfMessages: 10,
}));
for (const msg of r.Messages ?? []) {
  await handle(JSON.parse(msg.Body!));
  await sqs.send(new DeleteMessageCommand({ QueueUrl: queueUrl, ReceiptHandle: msg.ReceiptHandle! }));
}
```

### SQS FIFO (ordering)
```ts
await sqs.send(new SendMessageCommand({
  QueueUrl: fifoUrl, // .fifo 끝
  MessageBody: ...,
  MessageGroupId: order.userId,        // 같은 group 안 ordering
  MessageDeduplicationId: eventId,    // 5분 dedupe
}));
```

### NATS Core (fire-and-forget, 빠름)
```ts
nc.subscribe('telemetry.*', { callback: (err, msg) => handle(msg.data) });
nc.publish('telemetry.cpu', JSON.stringify({ host, cpu }));
```

영속 X — 받지 못하면 잃음. 모니터링 / IoT 같은 use case.

## 🤔 의사결정 기준
| 요구 | 추천 |
|---|---|
| 큰 처리량 + 영속 + replay | Kafka |
| 가벼운 + Quick start | NATS JetStream |
| Routing 복잡 (topic + header) | RabbitMQ |
| AWS only | SQS (+ EventBridge / SNS) |
| 매우 작은 (1대) | Redis Streams |
| Realtime fire-and-forget | NATS Core / Redis pub/sub |
| Lambda trigger | SQS / EventBridge |

## ❌ 안티패턴
- **모든 곳 Kafka**: 운영 부담 큼. 단순 큐 = SQS / RabbitMQ.
- **NATS Core 영속 가정**: subscribe 안 됐으면 잃음.
- **Prefetch 무제한 (RabbitMQ)**: 한 consumer 가 모두 가져감.
- **Visibility timeout 짧음 (SQS)**: 처리 중 다시 visible → 중복.
- **Manual ack 없음**: 처리 실패해도 ack — 메시지 잃음.
- **DLQ 없음**: 실패 메시지 영원 재시도 또는 잃음.
- **FIFO 가정 standard SQS**: ordering 보장 X.

## 🤖 LLM 활용 힌트
- 상황별 다름: 영속/replay = Kafka, 가벼움 = NATS, AWS = SQS, 복잡 routing = RabbitMQ.
- DLQ + idempotency 항상.
- prefetch / visibility 적정.

## 🔗 관련 문서
- [[Messaging_Kafka_Patterns]]
- [[Backend_Job_Queue_Patterns]]
- [[Messaging_Exactly_Once]]