Future Improvements 24/02/1999

* Use of a high performance network block device with software RAID-1
which would do away with the periodic mirroring and its associated caveats.

* Evaluation of the CODA distributed filesystem for a cluster
filesystem.

* Improved administrator notification and communication with network
management systems.

* Use of service watchdogs or agents which will monitor the health of a
particular service and initiate a node takeover when things look bad.

* Inclusion of a lower level and more robust heartbeat system which
could make use of the serial port or target mode SCSI instead of the more
complex network interface card.

* A GUI and setup scripts for cluster and node configuration.

* Improved scalability across more than two nodes.

* Dual sendmail queuing and delivery for concurrent operation.
