Performance Gap between Docker and K8s #47530
Replies: 5 comments 1 reply
-
|
帮顶 |
Beta Was this translation helpful? Give feedback.
-
|
up |
Beta Was this translation helpful? Give feedback.
-
|
up |
Beta Was this translation helpful? Give feedback.
-
|
standalone物理机docker是这种方式启动的？ https://milvus.io/docs/install_standalone-docker.md standalone也分为embeded etcd和external etcd， local storage和minio storage，不同方式虽然都是standalone，但运行方式不太一样。对于有external etcd和minio storage的standalone，会有3个容器。而milvus.io/docs/install_standalone-docker.md里面介绍的安装方式，standalone是embeded etcd和local storage模式，只有一个容器。 |
Beta Was this translation helpful? Give feedback.
-
|
另外，分别截取一段测试过程中的日志，standalone物理机docker的，以及K8s standalone的，我对比看一下为何latency会差这么多 |
Beta Was this translation helpful? Give feedback.
-
|
分布式比standalone慢是正常的，吞吐比较重要，latency没那么重要。不过你这个延迟明显高了，正常10-20ms |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
背景：团队内部搭的集群模式跑在K8s上，发现性能不达标，比业务方自己搭的standalone要慢很多
我们做了一个测试发现同样是standalone部署模式的情况下，物理机docker跑的比在K8s快很多。我们甚至将K8s的requests，limits等限制去掉（相当于跑裸机）也还是比物理机docker差很多。请问有遇到过类似情况么，有没有办法可以缩小这个差距？
P.S, 集群模式比standalone更慢一些...
Beta Was this translation helpful? Give feedback.
All reactions