Theodore Ts'o wrote:
Another major problem with this concept is that it lumps all I/O's into a single cgroup. So I/O's from pseudo filesystems (such as reading from /sys/kernel/debug/tracing/trace_pipe), networked file systems such as NFS, and I/O to various different block devices all get counted in a single per-cgroup limit. This doesn't seem terribly useful to me. Network resources and block resources are quite different, and counting pseudo file systems and ram disks makes no sense at all.
Yep, I know it. I've already mentioned about this as first planned improvement: |* Split bdi into several tiers and account them separately. For example: | hdd/ssd/usb/nfs. In complicated containerized environments that might be | different kinds of storages with different limits and billing. This is more | usefull that independent per-disk accounting and much easier to implement | because all per-tier structures are allocated before disk appearance. Accounting each BDI separately doesn't very useful too, so I've chosen something in the middle.
Regards, - Ted -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email:<a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx</a>
-- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>