Details
-
Improvement
-
Resolution: Won't Fix
-
Critical
-
5.5.0
Description
The current default memory allocation reserves a large chunk of the memory to the data service and 1/3 of the data service's amount to Analytics.
For a single node installation (e.g. for a developer testing the product) this partitioning of memory does not seem appropriate for an Analytics user as the Analytics service will potentially shadow the whole data of the data service - and require additional memory query processing. So the Analytics quota should approximately be the same as the KV quota to support this scenario.
For a multi-node cluster installation the data service should not run on the same machine as the Analytics service (to isolate analytical workloads from transactional workloads). So the Analytics service should use most of the memory of the node(s) that it is deployed on and not require any memory on other nodes.
We need
- to decide what kind of installation should be supported by the default allocation and
- determine a good allocation of memory for this installation.