Nordlicht Reisen GmbH, a fictional travel agency, has so far run its booking database on a SQL Server in its own data center in Hamburg: around 1 TB (assumption of the task), a .NET application in front of it, jobs at night for reporting tables, and on campaign days about ten times as much load as usual. The database is to move to Azure. The exam calls this: recommend a solution for relational data and recommend a service tier and a compute tier. There are three decisions in a row: the service, the tier and the compute form.
Prior knowledge. What Azure SQL is, the three forms (database, managed instance, VM), DTU versus vCore and the serverless offering are explained by K3 in lesson 3.6.1; Cosmos DB and PostgreSQL in overview by K3 in lesson 3.6.4. That is not repeated here. What is new are the limits that flip a choice. Scaling and protection follow in lesson 2.2, backup and high availability in lessons 3.1 and 3.4, the move itself in lesson 5.3.
The requirement
All numbers are assumptions of the task.
| Requirement | Source | Measure (assumption) |
|---|---|---|
| Booking database of about 1 TB, growth of about 300 GB a year | Operations | must grow over several years without a rebuild |
| Campaign days with ten times the load | Sales | compute must scale up fast and back down |
| Existing application without CLR, without queries across several databases | Development | 3 nightly jobs that can be rebuilt as scheduled functions or elastic jobs |
| Daily reports for sales | Sales | must not slow down bookings |
| Survive the failure of a zone | Management | zone-redundant |
| Frequent queries fast | Business unit | response time of the storage layer in the low millisecond range |
| Four development and test databases | Development | used only during the day |
| Existing SQL Server licenses with Software Assurance | Purchasing | use if possible (Azure Hybrid Benefit) |
| Little administration effort | IT management | two people in the data team |