On 9/3/23 15:08, Ing. Andrea vettori wrote:
Note: for a test search that retrieves only 10 documents, qtime is very low (2 msec) but the full request time to get javabin or json data is very slow (several seconds).
Lucene writes stored fields in a compressed format. Decompressing thousands of fields is going to be SLOW. I doubt that the compression is multi-threaded, so throwing more CPU power at it is probably not going to help.
I am not aware of any way to have Solr change the Lucene defaults for compression, but one thing you can do is turn on docValues and turn off stored for the majority of the fields, which requires reindexing from scratch. Lucene's docValues are NOT compressed, so retrieving lots of them is likely to be faster than stored. Fields using the solr.TextField class cannot have docValues.
Solr and Lucene are not really designed to deal with large numbers of fields. They are not the best choice for a primary data repository.
Thanks, Shawn
