Which option would ensure that the correct nodes are always available for the appropriate workload while meeting these requirements?
Use the writer endpoint for OLTP and the reader endpoint for the OLAP reporting workload.
Use automatic scaling for the Aurora Replica to have the appropriate number of replicas for the desired workload.
Create additional readers to cater to the different scenarios.
Use custom endpoints to satisfy the different workloads.
Explanations:
The writer endpoint is typically used for OLTP workloads (for both reads and writes), while the reader endpoint is used for OLAP reporting workloads (read-only). However, this option doesn’t allow specific control over directing OLAP workloads to the smaller nodes and OLTP workloads to the larger nodes, which is necessary in this case.
Automatic scaling for Aurora Replica does not control node sizes or allocate workloads to specific nodes. It scales based on demand, but it doesn’t ensure that the reporting workload will only use the small nodes or that the OLTP workload uses the large nodes as required.
Creating additional readers can help distribute read traffic, but it does not guarantee that the reporting workload will specifically use the two small nodes or that OLTP workloads will be directed to the large nodes. This option lacks fine-grained control over node allocation.
Using custom endpoints allows specific targeting of workloads to particular nodes. You can create custom endpoints for the two small nodes for the OLAP workload, and another for the four large nodes for the OLTP workload, ensuring that the appropriate resources are used for each workload.
My best guess is:
Use custom endpoints to satisfy the different workloads.