Oracle has several high availability and disaster recovery features supported on Azure virtual machines such as Golden Gate. I want to look specifically at the database-oriented products, Data Guard and Far Sync. Together these products offer synchronized local replication and asynchronous replication at a distance covering scenarios of localized hardware or data center failure and a region-wide failure.

Data guard provides high availability by using a synchronized replica of the primary database. It is the synchronization that ensures zero data loss. Synchronizing is acceptable within a data center or even within an Availability zone but isn't practical at any distance. The Far Sync functionality introduced in Oracle 12c, as the name suggests, takes care of keeping a distant standby copy of the primary database up to date. Oracle synchronizes databases by applying redo logs between databases. This is analogous to log shipping in SQL Server speak. A minimal configuration would be a primary database with a closely located Far Sync instance and a standby or replica database at a distance. The Far Sync instance contains no data files and exists purely for coordinating the asynchronous delivery of redo logs to the standby database. Oracle recommends using Availability Sets and premium or ultra SSD disks to provide as much localized protection as possible. As good as the promise of zero data loss at a distance is, the reality still runs into the same issues of any synchronous data replication solution. 

If the Far Sync instance is located in the same data center, it will fall victim to any data center wide failure. If it is located in another data center within the same region, synchronous replication delay may become an issue. Unless database responsiveness is an absolutely critical priority, having the primary database and Far Sync instance located within an Availability zone at different data centers is the safest option. In theory, this architecture provides a very low to zero RPO but says nothing about the RTO, time to get the standby at a distance database in write mode serving requests. An alternative, albeit more expensive solution, would be Data Guard replication between primary and standby databases within an Availability zone and Far Sync replication between the "local" standby and a standby located in a different region.