Before we get into specifics, I want to go over some definitions so we're all on the same page, as some of the terms sound like they could mean the same thing but are subtly different. High availability relates to a functioning system as opposed to the system's data. Redundancy, usually in the form of servers, but it can be disks or pretty much any infrastructure, is most often associated with high availability. When a server experiences a hardware failure, network traffic is re-routed to the mirror or backup server, often with users utterly unaware of the failure. The Recovery Time Objective is the time taken to switch between a primary server or system and its backup, most commonly referred to by the acronym RTO. In an ideal world, we would like this switch over time to be zero, or as it is usually expressed, 100% available. 99.99% availability sounds impressive but still translates into just under an hour's downtime annually, which could be very painful if it happened all at once during a system's maximum load. Implicit in high availability is load balancing and the ability for network traffic to be re-routed quickly and correctly in the event of a node failure.

There is the assumption that switching from a primary server to a backup means no loss of data. However, this is not the case for two reasons. 

1. Data may be lost during the failover process, even if that is a fraction of a second.

2. Suppose the data from the primary system isn't replicated to the backup system in real-time. In that case, data loss equals the difference between the backup system taking over and the last time data was updated to the backup system.

Recovery Point Object, RPO, is how much data you are willing to sacrifice in the case of a system outage and roughly relates to the frequency of data replication to back-ups or a backup system. You could say RTO relates to high availability and RPO is analogous to disaster recovery. These two concepts are related but subtly and crucially different. As we'll see, guaranteeing 100% availability with no data loss is an expensive proposition, especially when large amounts of data are involved.

The system data divide is also present when we think of the different elements within an SAP system. SAP utilizes Azure's Infrastructure as a service model, meaning from an application server perspective, virtual machine replication happens agnostically, that is, Azure doesn't care what's on the VM. The same cannot be said for data replication, where database-specific tools must replicate and backup data. 

High availability and disaster recovery are implemented in the geographical contexts of a data center, a group of data centers, called a zone, and Azure regions. Within a data center, you are looking to protect yourself from hardware failure, and implementing high availability and data replication is relatively inexpensive and efficient due to close proximity. As you extend system resilience to include situations where a whole data center or an Azure region goes down, costs go up, and replication efficiency goes down.