erofs

https://mp.weixin.qq.com/s/w7lIZxT9Wk6-zJr23oBDzA

https://cloud.redhat.com/blog/openshift-sandboxed-containers-network-performance

The sandbox receives one vhost process for each boot-time thread, but not for hot plugged threads. This is a known issue. Thus, it is only possible to increase the number of vhost processes by increasing default_vcpus, so if you determine that you need more network performance, you should use annotations to increase your VM size.  Setting this default higher may (or may not) improve pod performance

So as a general principle, do not set default_vcpus if you are also using at least a low CPU limit

If a pod has no CPU limit, increasing default_vcpus improves performance, at least up to the number of workers in the guest running uperf. However, with many threads, increasing the CPU limit does not help matters (at least up to the eight that we have tried), and indeed makes matters worse. So as a rule of thumb: to improve performance, use either a CPU limit or the default_vcpus annotation, but not both.

If there's no CPU limit, latency improves with increasing default_vcpus.

With a low CPU limit (less than the number of streams), however, latency degrades (this is even more apparent as the number of streams increases further).

With a large CPU limit, latency typically also improves, but is not better than with no CPU limit.