Message Queue Knowledge Base

Quick reference for message broker operations and advanced messaging patterns. Focuses on broker-level operations — for event-driven patterns, see eda-knowledge.

Broker Comparison

Feature	RabbitMQ	Apache Kafka	Amazon SQS	Redis Streams
Model	Message queue	Event log	Message queue	Event log
Ordering	Per-queue FIFO	Per-partition	Best-effort (FIFO available)	Per-stream
Retention	Until consumed	Time/size-based	4-14 days	Memory/size-based
Replay	No (once consumed)	Yes (offset seek)	No	Yes (ID-based)
Consumer Groups	Competing consumers	Native support	Not built-in	Native (XREADGROUP)
Throughput	~50K msg/s	~1M msg/s	~3K msg/s per queue	~100K msg/s
Latency	Sub-millisecond	Low milliseconds	10-100ms	Sub-millisecond
Protocol	AMQP 0-9-1	Custom (TCP)	HTTP/SQS API	RESP
Clustering	Quorum queues	Built-in (ZK/KRaft)	Managed	Redis Cluster
Best for	Task queues, RPC	Event streaming, logs	Serverless, AWS-native	Lightweight streaming

Message Delivery Guarantees

Guarantee	Description	Implementation	Trade-off
At-most-once	Message may be lost	Fire-and-forget, no ack	Fastest, data loss possible
At-least-once	Message delivered 1+ times	Ack after processing	Requires idempotent consumers
Exactly-once	Message processed exactly once	Transactional + deduplication	Slowest, most complex

Achieving At-Least-Once in PHP

// RabbitMQ: manual acknowledgment
$channel->basic_consume(
    queue: 'orders',
    no_ack: false,  // require explicit ack
    callback: function (AMQPMessage $msg) use ($channel): void {
        try {
            $this->handler->handle(json_decode($msg->getBody(), true));
            $channel->basic_ack($msg->getDeliveryTag());
        } catch (\Throwable $e) {
            $channel->basic_nack($msg->getDeliveryTag(), requeue: true);
        }
    },
);

Consumer Groups Overview

Broker	Mechanism	How It Works
RabbitMQ	Competing consumers	Multiple consumers on same queue; broker distributes round-robin
Kafka	Consumer groups	Partitions assigned to group members; each partition read by one consumer
Redis Streams	XREADGROUP	Consumer group tracks last delivered ID per consumer

Ordering Guarantees

Broker	Scope	Guarantee
RabbitMQ	Per-queue	Strict FIFO within single queue
RabbitMQ	Across queues	No ordering guarantee
Kafka	Per-partition	Strict ordering within partition
Kafka	Across partitions	No ordering guarantee
SQS Standard	Queue	Best-effort ordering
SQS FIFO	Message group	Strict FIFO within group
Redis Streams	Per-stream	Strict ordering by entry ID

When to Use Which Broker

Scenario	Recommended	Why
Task distribution (email, image processing)	RabbitMQ	Flexible routing, competing consumers
Event streaming / audit log	Kafka	Immutable log, replay, high throughput
Simple async in AWS	SQS	Managed, no infrastructure
Lightweight pub/sub with low latency	Redis Streams	Already have Redis, minimal overhead
RPC / request-reply	RabbitMQ	Built-in reply-to, correlation ID
CDC (Change Data Capture)	Kafka	Log compaction, connector ecosystem
Prioritized processing	RabbitMQ	Native priority queues
Cross-region replication	Kafka	MirrorMaker, built-in replication

Detection Patterns

# RabbitMQ usage
Grep: "AMQPChannel|PhpAmqpLib|bunny|php-amqplib" --glob "**/*.php"
Grep: "RABBITMQ_|AMQP_" --glob "**/.env*"

# Kafka usage
Grep: "RdKafka|kafka|KafkaConsumer|KafkaProducer" --glob "**/*.php"
Grep: "KAFKA_" --glob "**/.env*"

# SQS usage
Grep: "SqsClient|aws/aws-sdk.*sqs" --glob "**/*.php"
Grep: "SQS_|AWS_SQS" --glob "**/.env*"

# Redis Streams
Grep: "XADD|XREAD|XREADGROUP|XACK" --glob "**/*.php"
Grep: "xAdd|xRead|xReadGroup" --glob "**/*.php"

# Consumer patterns
Grep: "basic_consume|consume\(|poll\(" --glob "**/*.php"
Grep: "basic_ack|basic_nack|commitAsync|xAck" --glob "**/*.php"

# Dead letter configuration
Grep: "dead.letter|x-dead-letter|DLQ|deadLetter" --glob "**/*.php"

References

For detailed information, load these reference files:

• references/rabbitmq-advanced.md — Queue types, exchange topologies, clustering, monitoring, PHP patterns
• references/kafka-advanced.md — Partitioning, consumer groups, schema registry, exactly-once, PHP patterns