The best message brokers, according to hackers
19 message brokers recommended in Hacker News comments over the past year, ranked by how many separate people recommend each one.
- Most recommendedApache Kafka53 people recommend it
- Built to lastNATSowners report a median of 7.5 years
- Least regretCloudflare K210 recommend, 2 criticize
All message brokers
- Apache KafkaDurable, horizontally scalable messaging and event streaming platform that people reuse across many systems. Caveat: painful to operate and usually only worth it at large scale.53recommend
- NATSRecommended as a lightweight, low-latency messaging system that runs quietly for years. The most common caveat is its persistence and durability choices, which critics question.34recommend
- Amazon SQSCheap, reliable managed message queue that is simple to start with and needs little upkeep; critics warn of complex plumbing, timeouts and surprise usage costs.18recommend
- RedpandaKafka-compatible streaming platform that people find easier to run, quick to start, and lighter on resources than Kafka.17recommend
- RabbitMQOpen source broker people trust for performance under load, especially with Celery, though critics say it needs real expertise to configure and operate.13recommend
- PostgreSQL LISTEN/NOTIFYSimple built-in queue and pub/sub channel that people run in production without an extra service. Critics most often note no acknowledgements or retries, and doubt it scales under heavy load.13recommend
- Cloudflare K2People recommend K2 as a cheap, serverless, Kafka-compatible event stream for long retention; critics are reluctant to build on a non-standard platform and call the per-GB price steep.10recommend
- S2Developers use S2 for durable event streams backed by object storage, with a simple API and per-session streams for agents. Critics note the open-source S2-lite is single-node only.9recommend
- Apache IggyRust message streaming platform seen as a Kafka alternative with momentum; not yet distributed or fault tolerant.9recommend
- NATS JetStreamRecommended as a lightweight, easy-to-deploy persistence layer for NATS messaging, with high throughput and simple configuration. Critics report reliability bugs and data-storage failures.8recommend
- PgQuePostgres-based queue that gives durable messaging without running Kafka or RabbitMQ, built on an insert-only event table to avoid bloat. Critics doubt the zero-bloat claim.8recommend
- Cloudflare QueuesPeople recommend it for queues that handle retries and tracking of individual jobs, and for driving write costs down. Critics say its happy-path design lacks defensive handling.7recommend
- pgmqRecommended as a Postgres-native queue: just tables and a thin wrapper, with clients in many languages, so it is easy to inspect.6recommend
- Amazon KinesisAmazon's managed streaming service, valued because it removes much of the operational work of running streams. Critics say it adds many intermediate steps to data pipelines.6recommend
- PicoMQA single-binary, Kafka-like durable streaming service with an embedded admin dashboard, recommended for prototypes and simpler setups. The main caveat is that it cannot run serverless.6recommend
- TansuA Kafka-compatible message broker with a pluggable storage backend, recommended as a lightweight alternative to Kafka. Commenters say it is not yet proven for serious workloads.5recommend
- Apache PulsarFree open-source messaging system that supports the Kafka protocol and adds features like per-message acknowledgement. One critic notes it costs more in compute and operations than NATS.5recommend
- Google Cloud Pub/SubPeople recommend it as a managed messaging service that needs little upkeep, scales well, and handles batching, compression and duplicate-free delivery.5recommend
- WalrusRecommended as a high-performance, easier-to-run distributed message streaming engine written in Rust that combines log storage with Raft coordination. Critics question missing failure recovery and latency benchmarks.5recommend