We have an index containing about 1m records and somewhere around 40 fields.
The cluster is quite read-heavy and we query the data on multiple fields in different ways (exact match, less than/greater than, full-text search etc.), but we always need all query responses to be sorted on a particular date field.
The first thing that I'm asking that given the above facts, what are the performance considerations we should take into account about the cluster? E.g. caching settings.
The second question is that can we gain anything on query level from the fact that we always want all records in the response sorted on a particular field?
Also I would really like to know about your estimate on the average/worst case response time of the cluster while answering 1k queries (of the kind specified above) per second. (10 shards, 2 replicas each, on 7 nodes)
Thank you for your time and response in advance.