Configuring Network Bandwidth Management for Container

The network bandwidth for outgoing traffic a Container receives is controlled by two variables in the Container configuration file ( /etc/vz/conf/< CT_ID >.conf ): RATE and RATEBOUND .

Note: Container incoming traffic cannot be controlled in the current version of Parallels Server Bare Metal.

The RATE variable has the same format as TOTALRATE - < NIC >:< network_class >:< bandwidth > . This variable specifies the guaranteed outgoing traffic rate that the corresponding Container receives. This rate can be specified differently for different network classes and network adapters; use space to separate several rate descriptions.

Bandwidth values are specified in Kb/s. It is recommended to increase this value in 8 Kb/s chunks and to set it no lower than 8 Kb/s.

The RATEBOUND variable specifies whether the network bandwidth available to the Container for outgoing traffic is limited by the bandwidth specified in the RATE variable. The possible values of the RATEBOUND variable are yes and no ; the default is no . In this case the Container is allowed to take free bandwidth from the TOTALRATE pool.

The actual network bandwidth available to the Containers depends on the number of Containers and the total sum of the RATE values, and normally does not coincide with the bandwidth specified in their own RATE variables. If the RATEBOUND variable is set to yes , the Container bandwidth is limited by the value of the RATE variable.

If the Container configuration file does not specify any of these parameters, the values from the /etc/vz/vz.conf configuration file are taken. By default, Parallels Server Bare Metal does not set RATEBOUND , which corresponds to no , and RATE is set to eth0:1:8 .

The network bandwidth management in Parallels Server Bare Metal works in the following way. The bandwidth pool for a given network class (configurable through the TOTALRATE variable in the global configuration file) is divided among the Containers transmitting data proportionally to their RATE settings. If the total value of the RATE variables of all Containers transmitting data does not exceed the TOTALRATE value, each Container gets the bandwidth equal or greater than its RATE value (unless this Container has the RATEBOUND variable set to yes ). If the total value of the RATE variables of all Containers transmitting data exceeds the TOTALRATE value, each Container may get less than its RATE value.

The example below illustrates the scenario when there are two Containers, 101 and 102, which have RATEBOUND set to no , and Container 103 has RATEBOUND set to yes :

# grep ^RATE /etc/vz/conf/101.conf /etc/vz/conf/102.conf

RATE="eth0:1:8"

RATEBOUND="no"

RATE="eth0:1:8"

RATEBOUND="no"

# grep ^RATE /etc/vz/conf/103.conf

RATE="eth0:1:64"

RATEBOUND="yes"

With the default TOTALRATE of 4096 Kb/s, bandwidth pool will be distributed according to the following table:

After you have set up Container bandwidth settings, activate your changes as shown below:

# /etc/init.d/vz shaperrestart

Stopping shaping: Ok

Starting shaping: Ok

Set shaping on running Container: Ok

This command clears off all existing shaping settings and sets them again using the configuration files of running Containers.