I am setting up a new index and would like to use memory storage.
In order to configure its options I would like to understand what ends up
in small buffer/cache and what ends up in the large buffer/cache.
We have documents that are typically ~16kb each so if those count for small
I assume it is no use to allocate a lot of memory for e.g.
Also are these buffers and cache reserved on top of the actual indexed data?
As an example lets assume that we have an index of 3 GB size on a node that
has 8GB of total RAM (for simplicity let's assume a single node with no
replicas/shards). Any hints on your experience with memory storage or any
other best practices would be appreciated!