The service representing the function will be available a significant amount of the time but occasionally, the service may experience short periods of unavailable.
An example of such a quality of service would be a single service replicated on many machines. Many machines may go down but the service remains available. Of course, if all machines go down, service will not be available.
This level has one distinguishing characteristic of a known error free state. In a potentially automated fashion, the service in error is reset to the known error free state, usually an initial state, and restarted. The total distributed system continues to operate with what looks to the users as a minor event.
Highly available services are usually one or two point fault tolerant.