First, do you really need so many replicas? Try do decrease it to 1 or 0 and see if it helps (now you index your data 4 times, 1 for primary and same for each replica).
Second one, as in topic I mentioned disable _all field if you don't plan to use it.
If your data structure is predictable this will be great as you get full control of any field (type, analizer, doc_values and so on). Maybe disabling doc_values for fields which will never take part in aggregation would be good and give some speedup. Also exclude from indexing (index: no) if don't plan to search or filer on them.
You can also look at setting and discussion in ES indexing rate varies horribly