Why don't I have a GPU_IVF_FLAT in here

[gggdroa]

Is there an existing issue for this?

• I have searched the existing issues

Environment

- Milvus version:    image: milvusdb/milvus:v2.4.4-gpu
- Deployment mode(standalone or cluster):standalone 
- MQ type(rocksmq, pulsar or kafka):    
- SDK version(e.g. pymilvus v2.0.0rc2): pymilvus==2.4.4
- OS(Ubuntu or CentOS): Ubuntu 
- CPU/Memory: 
- GPU: 
- Others:

Current Behavior

1.GPU_IVF_FLAT cannot be used
2.How to optimize query speed

Expected Behavior

No response

Steps To Reproduce

No response

Milvus Log

No response

Anything else?

[gggdroa]

Query speed is slow how to optimize？
dim:1024; Data volume:110
use:GPU_IVF_FLAT "nlist": 5
search:"params": {"nprobe": 3}

[gggdroa]

Query speed is slow how to optimize？ dim:1024; Data volume:110 use:GPU_IVF_FLAT "nlist": 5 search:"params": {"nprobe": 3}

Query time 0.3 seconds

[xiaofan-luan]

if you have low latency requirment, we would recommend you to try Cagra index

[gggdroa]

if you have low latency requirment, we would recommend you to try Cagra index

Is the latency requirement a delay for adding data?

[gggdroa]

if you have low latency requirment, we would recommend you to try Cagra index

Is it a configuration problem? The server is not as fast as the local computer？GPU_IVF_FLAT is also used；
Query time 0.07 seconds

[xiaofan-luan]

70 ms not seems to be really slow. What is the expectation.
As far as I know, fastest GPU index takes 10-20 ms. HNSW could be faster if only latency is taken into consideration.
But GPU is definitely much better on throughput

[gggdroa]

70 ms not seems to be really slow. What is the expectation. As far as I know, fastest GPU index takes 10-20 ms. HNSW could be faster if only latency is taken into consideration. But GPU is definitely much better on throughput

Modified parameter： Query time 0.002s

Which is better in terms of concurrency performance, connections or Milvusclient? connections ？

[xiaofan-luan]

70 ms not seems to be really slow. What is the expectation. As far as I know, fastest GPU index takes 10-20 ms. HNSW could be faster if only latency is taken into consideration. But GPU is definitely much better on throughput

Modified parameter： Query time 0.002s

Which is better in terms of concurrency performance, connections or Milvusclient? connections ？

I think it's kind like similar.
I want to migrate from connections to milvusClient becasue it's sort of easier to understand

[gggdroa]

70 ms not seems to be really slow. What is the expectation. As far as I know, fastest GPU index takes 10-20 ms. HNSW could be faster if only latency is taken into consideration. But GPU is definitely much better on throughput

Modified parameter： Query time 0.002s
Which is better in terms of concurrency performance, connections or Milvusclient? connections ？

I think it's kind like similar. I want to migrate from connections to milvusClient becasue it's sort of easier to understand

How long is the viewable time Bounded? It's been in there for a while.

[xiaofan-luan]

by default it's 3 seconds but you can tune it

[stale]

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.