Apache Iggy Review (2026): Pricing, Features & Honest Verdict

TLDR

Apache Iggy is a persistent message streaming platform written in Rust that handles 1M+ messages per second per node at sub-millisecond latency, with QUIC, TCP, WebSocket and HTTP transports and SDKs for seven languages. It’s a leaner, single-binary alternative to Kafka. Best for: Rust-friendly teams needing fast streaming without JVM overhead. Price: Free (Apache 2.0, no paid plans). Rating: 7.9/10.

What is Apache Iggy?

Apache Iggy is a high-performance, persistent message streaming platform built from scratch in Rust. It does the same core job as Apache Kafka or Redpanda: producers write messages to streams, topics and partitions, consumers read them back (individually or in consumer groups), and everything is durably persisted to disk so nothing is lost on restart. The difference is the engineering underneath. Iggy is built on io_uring with a thread-per-core, shared-nothing architecture, so each CPU core runs its own pinned, NUMA-aware shard with no locks on the hot path, no garbage-collection pauses and no thread contention. It ships as a single binary with zero external dependencies, no ZooKeeper, no JVM, no sprawling cluster config just to run one node.

The project was created by Piotr Gankiewicz (GitHub handle spetz) in April 2023, originally under the MIT license and backed by his company LaserData Inc. He started it partly to learn Rust deeply and partly out of frustration with the operational weight of existing messaging systems he’d worked with, including RabbitMQ, ZeroMQ, Kafka and Aeron. Over its first two years it grew from a single-protocol experiment into a full streams-topics-partitions-segments architecture with multiple transports, a web UI and a maturing SDK ecosystem. In February 2025 the project was accepted into the Apache Incubator with 21 votes in favour and zero against, and the source was donated to the Apache Software Foundation under a Software Grant Agreement. It’s now an Apache 2.0 project and, at the time of writing, still in incubation, meaning it’s real and actively developed (with roughly 4,400 GitHub stars and a steady contributor base) but not yet a graduated top-level Apache project. That distinction matters: incubation signals promise and momentum, not yet the stability guarantees you’d get from a mature, graduated project like Kafka.

What Are Apache Iggy’s Key Features?

Raw Rust Performance

This is the headline. Published benchmarks show Iggy pushing throughput past 5,000 MB/s, roughly five million 1KB messages per second, with stable sub-millisecond P99 latencies on a single node given the right hardware. The io_uring plus thread-per-core design is what makes those numbers possible: each core owns its own shard, so there’s no lock contention, no cross-thread coordination on the hot path, and crucially no garbage-collection pauses to introduce tail-latency spikes. Custom zero-copy serialization means Iggy works directly with binary data rather than allocating and copying buffers, and messages are buffered in an in-memory journal before being flushed to disk via vectored I/O. For latency-sensitive workloads, this is genuinely a different class from JVM-based brokers, and it’s why the project frames itself around P99 latency rather than just average throughput.

Four Transport Protocols and Seven SDKs

Clients can connect over QUIC, TCP, WebSocket or plain HTTP, all with optional TLS. That flexibility is unusual, most brokers force one wire protocol. Official client libraries exist for Rust, C#, Java, Go, Python, Node.js and C++, so you’re not locked into a single-language ecosystem. There’s also a CLI and a web UI dashboard for administration.

Kafka-Style Semantics Without the Kafka Ops

Iggy implements the familiar model of streams, topics, partitions and segments, plus consumer groups with cooperative rebalancing for horizontal read scaling. It’s genuinely multi-tenant, with configurable message retention per stream and topic. But you get this without ZooKeeper, without a JVM to tune, and without the deployment ceremony Kafka is notorious for.

Security, Observability and an Ecosystem

Security is more complete than most young projects: TLS, per-stream and per-topic permissions, Personal Access Tokens, and optional AES-256-GCM encryption at rest. For observability there’s first-class OpenTelemetry logs and traces plus Prometheus metrics, so it drops straight into a modern monitoring stack without bolt-on exporters. The surrounding ecosystem is growing fast too, with connectors for PostgreSQL, MongoDB, Elasticsearch and Apache Iceberg that let you sink or source data without writing glue code, and even an MCP (Model Context Protocol) server for wiring streams directly into LLM workflows, which is a forward-looking touch you rarely see in a broker this young. Between the CLI, the web dashboard and these integrations, the day-to-day operator experience is more polished than the incubating label might suggest.

How Much Does Apache Iggy Cost?

Apache Iggy is 100% free and open source under the Apache License 2.0. There are no paid tiers, no hosted SaaS pricing, no per-message fees and no enterprise add-ons, you download the binary or pull the crate and run it yourself. This is a meaningful contrast with the commercial streaming world: Confluent Cloud (managed Kafka) starts with usage-based pricing that scales into hundreds or thousands of dollars a month at production volumes, and Redpanda’s managed cloud follows a similar per-throughput model. With Iggy your only cost is the infrastructure you run it on and the engineering time to operate it. That’s the classic open-source trade: zero license cost, but you own the ops. For a small team running high-volume event pipelines, that difference can amount to a five-figure annual saving versus a managed broker, provided you have the engineering capacity to self-host. And because it’s Apache 2.0 rather than a source-available or business-source license, there are no future clauses that could restrict commercial use, you can build a product on top of it without worrying about relicensing surprises down the line.

Who Is Apache Iggy Best For?

Iggy is aimed squarely at engineering teams that care about latency and infrastructure efficiency and are comfortable running their own streaming layer.

Use Apache Iggy if you want Kafka-style streaming with a fraction of the operational overhead, you’re latency-sensitive and want to escape JVM garbage-collection jitter, you’re already a Rust shop (or at least Rust-curious), you want a single-binary deploy for a self-hosted broker, or you value avoiding vendor lock-in and license fees. It’s also a great fit for high-throughput event pipelines where a single beefy node can carry the load.

Skip Apache Iggy if you need a proven, battle-tested system for mission-critical production today, you require multi-node clustering and automatic failover right now (Iggy currently runs as a single node; clustering based on Viewstamped Replication is planned but not shipped), you need a fully managed hosted service with an SLA, or your team wants the enormous connector and tooling ecosystem that mature Kafka already has.

Best Apache Iggy Alternatives

Iggy sits in a crowded streaming market. Here are the three most relevant alternatives depending on what you’re optimizing for.

Apache Kafka

The industry default for event streaming, with an unmatched ecosystem, Kafka Connect, and years of production hardening. It’s far more mature and scales to massive multi-node clusters, but it’s heavier to operate (JVM, historically ZooKeeper) and can suffer GC-related latency spikes. Kafka itself is free and open source; managed Confluent Cloud is usage-based and can run into the thousands per month at scale. Choose Kafka when maturity and ecosystem outweigh operational simplicity.

Redpanda

The closest philosophical cousin: a Kafka-API-compatible broker written in C++ with a thread-per-core, no-JVM, single-binary design and strong latency numbers. Redpanda is more established than Iggy and Kafka-compatible, which lowers migration friction. It’s source-available (BSL) for the core with a managed cloud on usage-based pricing. Pick Redpanda if you want Iggy’s efficiency ethos but need Kafka API compatibility and a commercial support option today.

NATS (JetStream)

A CNCF-graduated messaging system whose JetStream layer adds persistence and streaming. It’s extremely lightweight, easy to run, and great for cloud-native and edge scenarios. NATS is fully open source (Apache 2.0) and free. It’s less throughput-focused than Iggy for raw streaming benchmarks but wins on simplicity and a mature clustering story. Choose NATS if you want dead-simple, multi-node messaging that’s production-proven now.

Final Verdict: Is Apache Iggy Worth It?

Apache Iggy is one of the most exciting infrastructure projects we’ve come across in a while. The Rust plus io_uring plus thread-per-core foundation delivers benchmark numbers, five million messages a second at sub-millisecond P99 on a single node, that genuinely embarrass much of the JVM streaming world, and the single-binary, zero-dependency deployment is a breath of fresh air compared to standing up Kafka. That it’s fully Apache 2.0 licensed with no paid tiers makes it an easy thing to trial with zero budget risk.

The honest caveat is maturity. Iggy is still an Apache incubating project and currently runs as a single node, clustering is on the roadmap but not here yet, so for mission-critical, multi-region production you’ll want to wait or run it alongside a proven system. But for latency-sensitive pipelines that fit on one powerful node, self-hosted event streaming where you want to drop your license bill to zero, or any Rust-leaning team that wants a modern broker, Iggy is well worth a serious pilot. Watch this one closely: if the clustering work lands as planned, it could become a real Kafka challenger.