What do you mean by "work localy without any interruption"?
Do you mean that users will connect to another server for the DB, but if that DB is unavailable, it will connect to the server's own DB?
That doesnt make sense, since you might as well implement the whole database without using the external server.
Otherwise, you'll have to batch update either server consistently.
i.e. when the local DB is used, once the remote DB is online, it is updated to match the local DB.
Under normal operation on the remote DB, the local DB must be updated to ensure data integrity in case the remote DB fails.
Conversely, you could mean that users work locally on their own computer.
The problem with that is that how are they going to update their info unless there is a location to store it? And if such a location exists, it serves as a database.