lucene has bloomfilter codec for low frequency field (like primary keys), why not use bloom filter codec as default codec for uid field in elasticsearch?
In the past, we used bloom filters on the UID field. But improvements in Lucene and how we use Lucene's TermsEnum made the usage of blooms obsolete. We were able to remove the bloom filters, keep the same query performance and reduce memory usage by not requiring the extra bloom data structure.
You can read some of the history here:
This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.