are there online resources that summarize how is the performance of
ElasticSearch/Lucene affected by various factors, like the number of
documents indexed, number of replicas used, rate of indexing new
documents, number of facets requested...? Some trench stories, hints,
insights, big-O overviews.
In particular, I am considering ES for an app and before committing,
I'd like to know:
The delay between "bulk indexing docs"->"these docs available for
search" is reasonably fast, but I only have a few testing docs in my
index. How will this delay be affected as the index grows bigger?
Will this lag increase as I add node replicas?
As I increase the number of shards?
What are the ramifications of lowering the "refresh" parameter (I
understand the default is 1 second)?
Does it make sense to turn off "refresh", and call refresh manually
after each bulk-index batch instead?