Description
Here is the use case.
Couchbase is deployed in a Cloud provider (GCP/AWS/Azure...) on 3 AZ (With 2 replicas in 1 bucket).
So that we get a copy of the documents in each AZ.
Application is deployed on 3 AZ as well.
Write traffic is low but Read traffic would be high.
Network traffic between AZ has a significant cost and goal is to reduce traffic cross AZ (between application and Couchbase nodes)
To do so, we could do the following:
=> Retrieve documents from the preferred/closest Group/AZ (could be replica) instead of master being 2/3rd of the time in another AZ.
=> SDK could be configured to provide preferred server group. (application would own the affinity)
=> All nodes in the preferred server group would be sort of local cluster for read operations to the application nodes in the same AZ.