Question about query performance #37823
Unanswered
artesby
asked this question in
Q&A and General discussion
Replies: 2 comments 2 replies
-
usually it takes 10-100 ms to do search. Questions:
We would definitely like to offer further help. Feel free to reach me out at [email protected] |
Beta Was this translation helpful? Give feedback.
0 replies
-
Thanks for your reply.
I've done some more diagnostics, here the observations:
proxy:
query:
this sounds to me that there is some issue with mix coord, isnt it? it has warnings as well, but timings are not matching with search attempts.
|
Beta Was this translation helpful? Give feedback.
2 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Hello.
Our milvus cluster have about 20 Upsert RPS into 3 collections, avg vector size is 768.
Current collection size is ~5 million each.
Each collection uses FLAT index with consistency "bounded".
When we started quering these collections we noticed that query latency is 60s.
We tried to stop inserting (~30min) and restarted all components, without success.
Finally, setting search consistency to "strong" in pymilvus client reduced latency to ~500ms.
However, i am still wondering, what could make bounded consistency work so long? And what is expected search latency on 5-10mill FLAT index?
Beta Was this translation helpful? Give feedback.
All reactions