sam-herman commented on issue #14681: URL: https://github.com/apache/lucene/issues/14681#issuecomment-3005729455
> [@sam-herman](https://github.com/sam-herman) flush is indeed single threaded. > > I was mentioning this in the context of merge. Which is the only case in which its necessary as the initial size of segments is tiny when compared with merged segments. @benwtrent Thank you for confirming my understanding! Whether doing this through jVector codec or if we want to add it to default Lucene HNSW codec, I think multi threaded construction of graph is quite useful for a number of use cases: 1. **Agility** - Testing and experimentation is much faster when we want to isolate learning regarding graph codec over 1B+ vectors on a single graph. 2. **Offline build** - We can't always assume graph buildup is done using pre-existing smaller graphs, there are scenarios in which customers would like to build large (1B+ vector count) graphs offline directly from the data. -- 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
