Query with has_parent vs denormalized with lots of updates

Maybe some of you have some experience with this. Currently my index has a has_parent relationship of 1 parent to many children (roughly 200 children on average per parent).

This unfortunately leads queries to be slow, as I always need to query with has_parent and a size of 1.

Would I be better off denormalizing the index and updating ~300,000 docs every 2 minutes (instead of roughly 2,000 parents)? To note, this is an index that is also adding roughly ~5,000 new documents a second, which is what is triggering the warmers to cache the parent-child relationships.

Hi @hello,

Unfortunately it is a big depends although updating 300k docs seems to be worse scenario.
I would encourage you to create a benchmark with both approaches and observe which perform better.

Maybe rally can help you with this task.


This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.