Table of content

Linux Kernel Scheduler

A kernel scheduler is a unit of the kernel that determines the most suitable process out of all runnable processes to execute next; it allocates processor time between the runnable processes on a system. A runnable process is one which is waiting only for CPU time, it’s ready to be executed.

The scheduler forms the core of multitasking in Linux, using a priority-based scheduling algorithm to choose between the runnable processes in the system. It ranks processes based on the most deserving as well as the need for CPU time.

Process Priority and Nice Value

The kernel stores a great deal of information about processes including process priority which is simply the scheduling priority attached to a process. Processes with a higher priority will be executed before those with a lower priority, while processes with the same priority are scheduled one after the next, repeatedly.

There are a total of 140 priorities and two distinct priority ranges implemented in Linux. The first one is a nice value (niceness) which ranges from -20 (highest priority value) to 19 (lowest priority value) and the default is 0, this is what we will uncover in this guide. The other is the real-time priority, which ranges from 1 to 99 by default, then 100 to 139 are meant for user-space.

Check Nice Value of Linux Processes

To view processes nice value with ps command in user-defined format (here the NI the column shows the niceness of processes).

$ ps -eo pid,ppid,ni,comm

Alternatively, you can use top or htop utilities to view Linux processes nice values as shown.

Difference Between PR or PRI and NI

From the top and htop outputs above, you’ll notice that there is a column called PR and PRI receptively which shows the priority of a process.

This, therefore, means that:

  • NI: is the nice value, which is a user-space concept, while
  • PR or PRI`: is the process’s actual priority, as seen by the Linux kernel.

How To Calculate PR or PRI Values

Total number of priorities = 140
Real time priority range(PR or PRI):  0 to 99
User space priority range: 100 to 139

Nice value range (NI): -20 to 19

PR = 20 + NI
PR = 20 + (-20 to + 19)
PR = 20 + -20  to 20 + 19
PR = 0 to 39 which is same as 100 to 139.

But if you see a rt rather than a number as shown in the screenshot below, it basically means the process is running under real-time scheduling priority.

Run A Command with a Given Nice Value

Here, we will look at how to prioritize the CPU usage of a program or command. If you have a very CPU-intensive program or task, but you also understand that it might take a long time to complete, you can set it a high or favorable priority using the nice command.

The syntax is as follows:

$ nice -n niceness-value [command args]
# or
$ nice -niceness-value [command args]   #it’s confusing for negative values
# or
$ nice --adjustment=niceness-value [command args]

Important:

  • If no value is provided, nice sets a priority of 10 by default.
  • A command or program run without nice defaults to a priority of zero.
  • Only root can run a command or program with increased or high priority.
  • Normal users can only run a command or program with low priority.

For example, instead of starting a program or command with the default priority, you can start it with a specific priority using following nice command.

$ sudo nice -n 5 tar -czf backup.tar.gz ./Documents/*
# or
$ sudo nice --adjustment=5 tar -czf backup.tar.gz ./Documents/*

You can also use the third method which is a little confusing especially for negative niceness values.

$ sudo nice -5 tar -czf backup.tar.gz  ./Documents/*

renice

Change the Scheduling Priority of a Process

As we mentioned before, Linux allows dynamic priority-based scheduling. Therefore, if a program is already running, you can change its priority with the renice command in this form:

$ renice -n -12 -p 1055
$ renice -n -2 -u apache

The niceness of the process with PID 1055 is now -12 and for all processes owned by user apache is -2.

Still using this output, you can see the formula PR = 20 + NI stands,

PR for ts3server = 20 + -12 = 8
PR for apache processes = 20 + -2 = 18

Any changes you make with renice command to a user’s processes nice values are only applicable until the next reboot. To set permanent default values, read the next section.

Set Default Nice Value Of a Specific Users Processes

You can set the default nice value of a particular user or group in the /etc/security/limits.conf file. Its primary function is to define the resource limits for the users logged in via PAM.

The syntax for defining a limit for a user is as follows (and the possible values of the various columns are explained in the file):

#<domain>   <type>  <item>  <value>

Now use the syntax below where hard – means enforcing hard links and soft means – enforcing the soft limits.

<username>  <hard|soft>  priority  <nice value>

Alternatively, create a file under /etc/security/limits.d/ which overrides settings in the main file above, and these files are read in alphabetical order.

Start by creating the file /etc/security/limits.d/franzmusterman-priority.conf for user franzmusterman:

franzmusterman  hard  priority  10

Save and close the file. From now on, any process owned by franzmusterman will have a nice value of 10 and PR of 30.