Anubis

Anubis is a partition-aware tuple-space that can be used to share information reliably between hosts on a single site. It uses multicast IP and other techniques to share information across systems. Each host is notified of detected partition change events, as well as tuple insertion and removal.

What that means is that you can use it to reliably share information between machines on a single site/subnet supporting multicast IP. Clearly, it is not designed for use over long distances. Add a fact into the space, wait a heartbeat and everyone else knows it -unless a partition event has occurred, implying that at least one machine is isolated or has failed.

To test partition behaviour, the Anubis Partition Manager lets you view the current state of the cluster, and let you simulate network partitions, to see how the other machines in the network can handle the event, and tune heartbeat timeouts.

• Anubis Design Presentation PPT
• Anubis User Guide PDF
• The Anubis Service PDF
• A Distributed State Monitoring Service for Adaptive Application Management PDF