Nordlicht Reisen GmbH is moving its booking database, application servers and files to Azure. A backup belongs to each of these decisions, but which one depends on what may be lost and how long the return may take. This lesson chooses the backup for compute, databases and unstructured data.
Prior knowledge. Agreeing RTO and RPO with the business owner, point-in-time restore and failover groups of Azure SQL, and the restore test are taught by K4, lesson 4.7.5; backup and restore from the development point of view by K3, lesson 3.6.5. Vaults, policies and backing up a VM are set up in AZ-104, lesson 5.5. None of that is repeated here; lesson 3.2 of this course covers recovery after a disaster. What is new here is the choice: which service backs up which data and which limit flips the choice.
The requirement
All numbers are assumptions of the task, not measured.
| No. | Data | RPO | RTO | Further requirement |
|---|---|---|---|---|
| S1 | Booking database (about 1 TB) | 15 minutes | 4 hours | Errors often go unnoticed for weeks; keep booking data for ten years |
| S2 | Application servers, six Windows VMs in Azure | 24 hours | 4 hours | An attacker with administrator rights must not be able to delete backups |
| S3 | Two Windows servers in Hamburg that stay | 24 hours | 8 hours | Files, folders and system state |
| S4 | Office shares and contracts (lesson 2.4 of this course) | 24 hours | 8 hours | Contracts are additionally immutable |
RPO and RTO are decisions of the business, not of technology; K4, lesson 4.7.5 shows how to agree on them. Here they are taken as given.