Provides event notification services.
The system was designed to avoid vendor lock-in and use in business oriented 1:M event notification such as micro service-based business architectures.
The producers send messages to named "namespace/queues" which can optionally be partitioned by ShardKey.
The message sequencing by time is preserved.
Consumers poll/read the messages from distributed queue nodes in cluster region. The consumer keeps track what it has already read. A consumer can restart processing. In this regard, the design is similar to Apache Kafka.
The system is multi-master within cluster origin AND cross-region. An optional cross-origin replication service allows for multi-region multi-master replication however the ordering of produced messages may be out-of-sequence between messages produced in different regions.
The system has the following traits:
- Unidirectional append only distributed, fault tolerant commit log of events
- An event is a
byte[]with content type and headers - Logical event sequence preserved within origin (cluster region)
- Leader-less design, multi-master per cluster region (origin)
- Fault tolerant with 1+ queue node copies in the cluster
- Avoids vendor lock-in (when using default server implementation)
- Server broker backend plug-and-play, default uses Mongo Db collections, can use MySql, Kafka, etc.
- Simple implementation, very few parameters to configure
- Multi-region /multi-master option with replication agent