jpountz commented on PR #14412: URL: https://github.com/apache/lucene/pull/14412#issuecomment-2764454882
> I am wondering if we should also make maxRamBytesUsed dynamic as well. What is your use-case for tuning it? Related to my comment on the issue linked to this PR, I worry a bit about optimizing too much for query caching vs. making the query cache less relevant. To expand a bit more on this, for "good" queries, like term queries or FieldExistQuery that produce good iterators with a low up-front cost, memory would be better spent on the filesystem cache than on the query cache. Boolean queries recently had big improvements that makes the query cache less relevant. (annotations HS, HW and HX at https://benchmarks.mikemccandless.com/CountAndHighHigh.html and https://benchmarks.mikemccandless.com/CountOrHighHigh.html). So that makes the query cache mostly relevant for queries that have a high up-front cost, like point queries and multi-term queries. Rather than investing in the query cache for these queries, I would rather invest in making them behave better, e.g. recent work on vectorizing this up-front cost for `PointRangeQuery` so that it's less likely to be the bottleneck of query evaluation, or `IndexOrDocValuesQuery` to skip this up-front cost entirely when it's hurting more than helping. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For additional commands, e-mail: issues-h...@lucene.apache.org