mappedbus.io

What It Is

Mappedbus is a high-throughput, low-latency Java message bus library for IPC between multiple Java processes/JVMs. It uses a memory-mapped file or shared memory as the transport medium. The design is inspired by Java Chronicle, but places more emphasis on efficiently supporting multiple writers, making it suitable for multi-process scenarios where message production order matters.

Core Capabilities

Functionally, it supports object-based messages as well as byte array/raw data messages. The example API mainly revolves around MappedBusReader and MappedBusWriter. Writers can write objects or buffers, while readers poll next() and then read the message type and content. Internally, it handles multi-writer synchronization through mechanisms such as an 8-byte limit at the beginning of the file, per-record status flags, fetch-and-add, volatile, and compare-and-swap. It also provides timeout and rollback logic for uncommitted records caused by writer crashes. In terms of performance, the published benchmark reports around 14.13 million messages/second with one producer and one consumer on an i7-4558U laptop, with average read/write latency of about 70ns. After startup, it no longer creates objects, reducing GC impact.

Pricing and Deployment

The crawled content does not mention commercial pricing or paid plans. The project provides a mappedbus.jar download, and it can also be cloned and built from source with ant. It is essentially an embedded Java library rather than a SaaS service, so it can naturally be self-hosted inside local or server-side applications. The documentation includes Javadoc, examples, and performance tests, but Maven/Gradle coordinates and license information were not found.

Pros and Cons

Its strengths are a clear positioning, explicit performance goals, a simple API, and relatively concrete low-level design notes for multi-writer support and crash rollback. The drawbacks are also obvious: platform compatibility is limited. The text explicitly says the approach appears to work well on Linux x86 with Oracle JVM 1.8, but may not work on all platforms. Production operations documentation, community ecosystem, commercial support, and integration information are also limited.

Who It’s For and Access from China

It is a good fit for Java teams building low-latency messaging channels between multiple JVMs on the same machine, high-frequency market data/trading distribution systems, or shared-memory IPC prototypes. It is less suitable for teams that want something ready out of the box, strong cross-platform guarantees, or enterprise support. Access from China cannot be determined from the provided content alone. Alternatives worth evaluating include Chronicle Queue, Aeron, Disruptor, ZeroMQ, or Netty-based IPC solutions.