Indexing performance of high cardinality field


(Shahar Mor) #1

Hi,

Considering i have a long typed field with very high cardinality,
is it possible that it will affect CPU performance a lot for the indexing of it?

I am not using it for filtering, only for the aggregation results.

Should i consider change its mapping to index: no? Have anyone seen a performance boost for indexing from it?


(Boaz Leskes) #2

of course it all depends. A long field typically doesn't carry much overhead but if you have a Iot of them and you don't need it for searching, you can disable indexing and see what happens. Of course, if you change your mind and you'd like to search it after all, you'd need to reindex.


(Shahar Mor) #3

But does the cardinality of the field matters? Or the number of events with this field?

Also, if i disable indexing for the field, does it mean i won't be able to search/filter by it, or just that it won't be fast?


(Boaz Leskes) #4

Both. Each value needs to be index. Note though, the total unique values in an index tends to matter less here. It's the total values to be indexed per second.

Also, if i disable indexing for the field, does it mean i won't be able to search/filter by it, or just that it won't be fast?

If you don't index it, you won't be able use it for filtering/searching and sorting.


(system) #5