QoS mechanism in Linux-cgroup (2)

This article focuses on how cgroup controls the QoS of the blkio sub-system. First, let's look at the settings

Blkio. Weight blkio. weight_device my understanding is that blkio. weight is the weight of the entire Io subsystem, while blkio. weight_device is the weight of a specific major: minor device.

Blkio. Throttle. Limit blkio. Throttle. write_bps_device blkio. Throttle. Limit blkio. Throttle. write_iops_device the maximum number of read/write BPS for specific major: minor devices.

The other blkio settings are the perf report data of the current blkio sub-system. Most of the data is similar to the iostat-x attribute and will not be repeated.

# Mount-T cgroup-O blkio NONE/cgroup/blkio

# Mkdir-P/cgroup/blkio/tinker

# Mkdir-P/cgroup/blkio/tailor

# Echo 800>/cgroup/blkio/Tinker/blkio. Weight

# Echo 200>/cgroup/blkio/tailor/blkio. Weight

Next we will use dd to write a large file and put the two processes into different cgroups. We can see that blkio changes to Io QoS by comparing the changes before and after.

Before using cgroup, we can see that I/O bandwidth is evenly allocated.

After cgroup is used, the ratio is basically 800: 200.

TIPS: cgroups of a virtual machine that uses PVM cannot work normally. I wonder if it is related to xenlinux.