A broker can only scale so far. The limitations of a broker-based messaging service are becoming apparent as Openstack deployments are pushed to ever higher scales and wider distributions. An alternative to the broker-based message bus is needed.
This presentation will introduce that alternative. With the Newton release, the Oslo.Messaging library now includes support for a highly scalable, distributed, and fault tolerant brokerless message bus based on message routing rather than queueing. With this new technology one can achieve high availability through redundancy rather than clustering. This message bus is optimized for operation across geographically distant sites, allowing for cloud distribution not possible with the single or federated broker approach.
Attendees will learn about an innovative new alternative to the broker-centric model used for inter-project messaging. In this session we will demonstrate a distributed messaging backend based on a mesh of high performance message routers. This demonstration will show how availability and fault tolerance can be greatly increased for large scale geographically distributed deployments.