[Bug]: [perf][Recall] Milvus diskann's recall of the data set in glove continues to decrease

[jingkl]

Is there an existing issue for this?

• [X] I have searched the existing issues

Environment

- Milvus version:2.2.0/master
- Deployment mode(standalone or cluster):
- MQ type(rocksmq, pulsar or kafka):    
- SDK version(e.g. pymilvus v2.0.0rc2):
- OS(Ubuntu or CentOS): 
- CPU/Memory: 
- GPU: 
- Others:

Current Behavior

glove-200 dataset-2.2.0 branch image： 2.2.0-20230429-546d3179 cluster

standalone image： 2.2.0-20230428-67dd70f1

glove-200 dataset-master ：master-20230505-6a48de34

sift 128d dataset-master master-20230504-a76e9d20

Expected Behavior

No response

Steps To Reproduce

1. create a collection
2. build disk and index
3，insert glove-200d data
4. build index again
5. load collection
6. search

Milvus Log

No response

Anything else?

No response

[liliu-z]

From this info, it is hard for me find out the potential issues. Can I have the commit ID of good/bad version? Thanks.

[liliu-z]

recall jump between bbc21fe8 and 67dd70, Knowhere versions are 1.3.10 and 1.3.12

[liliu-z]

/assign @hhy3

[hhy3]

Is there any way to reproduce the results? The 0.998 recall for glove-200 dataset seems astonishingly high.

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions. Rotten issues close after 30d of inactivity. Reopen the issue with /reopen.