minio / minio

The Object Store for AI Data Infrastructure

Home Page:https://min.io/download

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

minio single-node mode met performance bottleneck

sunwuyan opened this issue · comments

Hi,
since there is no storage class in the kubernetes cluster, but the same network file storage service is mounted on all nodes in the cluster, minio is deployed in the kubernetes cluster in single-node mode. The minio version is RELEASE.2023-07 -11T21-29-34, this minio service has only one bucket, and this bucket contains 200T data. When there are a large number of readers and writers (large files are read and written, and a large number of small files are written), the minio query response time is very long. , I allocated a lot of CPU and memory to this minio pod, but only less than half of it was actually used. The fio test showed that the bandwidth and iops of network file storage are acceptable, and the iops is 13w. I don't know what the bottleneck of minio is at this time. where? How to optimize it? Thanks!

Hi, since there is no storage class in the kubernetes cluster, but the same network file storage service is mounted on all nodes in the cluster, minio is deployed in the kubernetes cluster in single-node mode. The minio version is RELEASE.2023-07 -11T21-29-34, this minio service has only one bucket, and this bucket contains 200T data. When there are a large number of readers and writers (large files are read and written, and a large number of small files are written), the minio query response time is very long. , I allocated a lot of CPU and memory to this minio pod, but only less than half of it was actually used. The fio test showed that the bandwidth and iops of network file storage are acceptable, and the iops is 13w. I don't know what the bottleneck of minio is at this time. where? How to optimize it? Thanks!

You are perfect case to sign-up to our subscription https://min.io/pricing

Hi, since there is no storage class in the kubernetes cluster, but the same network file storage service is mounted on all nodes in the cluster, minio is deployed in the kubernetes cluster in single-node mode. The minio version is RELEASE.2023-07 -11T21-29-34, this minio service has only one bucket, and this bucket contains 200T data. When there are a large number of readers and writers (large files are read and written, and a large number of small files are written), the minio query response time is very long. , I allocated a lot of CPU and memory to this minio pod, but only less than half of it was actually used. The fio test showed that the bandwidth and iops of network file storage are acceptable, and the iops is 13w. I don't know what the bottleneck of minio is at this time. where? How to optimize it? Thanks!

You are perfect case to sign-up to our subscription https://min.io/pricing

Why closed my issue?

We do not perform community performance analysis.

We do not perform community performance analysis.

ok,tks!