Process Scheduling & States in Linux

Process States in Linux

A process in Linux can be in one of the following states:

1. Running (R) – The process is either executing on the CPU or ready to run.
2. Sleeping (S & D)
   • Interruptible Sleep (S) – The process is waiting for an event and can be woken up by signals.
   • Uninterruptible Sleep (D) – The process is waiting for I/O and cannot be interrupted.
3. Stopped (T) – The process is paused, usually due to signals like SIGSTOP or SIGTSTP.
4. Zombie (Z) – The process has completed execution but still has an entry in the process table because its parent has not retrieved its exit status.
5. Idle (I) – Used by kernel threads that are not actively doing any work.

You can view process states using the ps command:

   ps aux

Process Scheduling in Linux

Linux uses a priority-based scheduling mechanism. The scheduler decides which process gets CPU time based on scheduling policies and priorities.

Types of Scheduling Policies

Linux supports several scheduling policies, divided into real-time and normal policies.

1. Normal Scheduling Policies

These are used for standard user-space processes:

• SCHED_OTHER: Default time-sharing policy, using the Completely Fair Scheduler (CFS).
• SCHED_BATCH: Used for batch processing tasks that do not need interactivity.
• SCHED_IDLE: Lowest priority for background tasks.

2. Real-Time Scheduling Policies

These have higher priority and are used for time-sensitive tasks:

• SCHED_FIFO (First In, First Out): A real-time scheduling policy where the highest-priority process runs until it voluntarily yields the CPU.
• SCHED_RR (Round Robin): Similar to FIFO but includes time slices to ensure fairness.

You can check or modify the scheduling policy of a process using chrt:

chrt -p

Linux Scheduler: Completely Fair Scheduler (CFS)

Linux primarily uses the Completely Fair Scheduler (CFS), which assigns CPU time proportionally based on a process’s priority.

Concepts of CFS:

• vruntime (Virtual Runtime): Determines the amount of time a process has run; lower vruntime processes get CPU access sooner.
• Nice Value: Ranges from -20 (highest priority) to 19 (lowest priority), affecting scheduling.

You can change a process’s priority using nice and renice:

nice -n 10 ./myprogram

renice -n -5 -p

Monitoring Process Scheduling

Use the following commands to inspect and manage scheduling:

• top – Displays real-time CPU usage and process priorities.
• htop – An interactive process viewer.
• schedtool – Displays and sets process scheduling policies.

Example to check a process’s priority:

     ps -eo pid,comm,pri,nice