We've one usecase to implement where we want to maintain a relationship between two types of documents. We tried to put in the same document but the document size exceeding 200kb. So we're exploring parent-child relation model, where 1parent will map to 1 child. Read through the documentation regarding the performance issue when using the joins, but want to check any room for more improvement here.
Es version - 7.12
Cluster - 250+ mn document parent & 250+ mn child
200+ shards, writes are not frequent, mostly read-only cluster.
After indexing, field data is size 2gb, single query with has_child is taking almost 2.5 - 5 secs.
_stats/fielddata?human&fields=join_field#parent
primaries: {
fielddata:{
memory_size: "2.1gb",
memory_size_in_bytes: 2309747432
}
}
Is there a way to improve these metrics?