Scaling Milvus cluster #34370
Unanswered
nairan-deshaw
asked this question in
Q&A and General discussion
Replies: 2 comments 1 reply
-
|
Beta Was this translation helpful? Give feedback.
0 replies
-
@xiaofan-luan question around the querynodes: Is it recommended to have multiple querynodes with smaller memory or less querynodes with larger memory? |
Beta Was this translation helpful? Give feedback.
1 reply
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
We currently have a cluster deployment of Milvus on K8s with close to 1TB in memory. There are about 16 query nodes each with 40G of memory. We're looking at data in the range of 250M vectors of 1536 dimension. Currently it uses IVF_FLAT indexing.
The main concerns are the slow partition load and release times before we run the queries. We are looking to through some more resources at the cluster and up the memory to about 3.5TB. What would be some recommendations for the number of query nodes and the memory assigned to each node. I've taken a look at the sizing tool but we also would like some recommendations here.
Some more related questions on this aspect:
Beta Was this translation helpful? Give feedback.
All reactions