4 ms·
Thanks for your feedback. We should have emphasized that we are benchmarking Qdrant in local mode. Based on our analysis of the local mode, Qdrant persists the
by woshiwan 3y ago
Thanks for your feedback. We should have emphasized that we are benchmarking Qdrant in local mode. Based on our analysis of the local mode, Qdrant persists the index in this manner: at the directory where qdrant persists index /data/workspace/vector-benchmark/trainQDRANTIndex
> du -sh storage.sqlite
2.0G storage.sqlite
--------------------------------------
> sqlite3 storage.sqlite
--------------------------------------
> sqlite> PRAGMA table_info(points);
0|id|TEXT|0||1
1|point|BLOB|0||0
--------------------------------------
> sqlite> select count(*) from points;
1000000
--------------------------------------
> sqlite> select * from points limit 1;
gARLAS4=|��,
--------------------------------------
We have updated the post to clarify that Qdrant is being evaluated in local mode.
- timvisee 3y agoI think it shouldn't be in this article at all. You're comparing a tool that is not optimized nor efficient in any way against others that are. That while Qdrant has a very well optimized offering as well. This is comparing a walker against modern cars and paints a bad picture.