Always Availability

Conventional High Availability (HA) systems always seem to be struggling to achieve higher reliability, approaching this goal as a limit. That's because they share a basic design flaw: the shared storage approach. An architecture that requires two systems to share a single storage system inherently creates several problems:

The storage system becomes a single point of failure. Backup storage becomes a structural requirement for redundancy.
Only one system can read or write to a given file at a given time. The other system is stand-by mode and cannot perform read or write functions.
When the primary system fails, the stand-by system has to perform a disaster recovery procedure. First it checks the integrity of file system on the storage system, then it fixes any problems it detects, then it mounts the file system on the storage system. Only then can it perform read or write functions. This can be very time-consuming.
The HA system and its storage systems, primary and backup, have to reside in a single location. Earthquakes, fires, floods, and other disasters can shut shut the HA systems down completely.

The Mirror File System uses a completely different approach. It enables the nodes of a clustering system to be separated by hundreds miles or more, with each node freely accessing the files on its own storage system. Files and operations are mirrored, rather than shared, so when a file is updated on one node, the same file on the other node is also updated, in real time. When one node crashes or shuts down, the other node remains up and running. No disaster recovery procedure is needed, and the data remains Always Available. That is the point.