Bo2SS

Bo2SS

HomeArchives

4 Advanced Process Management

#计算机#操作系统#Linux#海贼班82
AI Translation
This post is translated from Chinese into English through AI.View Original
AI-generated summary
The document discusses advanced process management in operating systems, focusing on process scheduling. It explains the roles of ready, blocked, and running states in a process's lifecycle, and introduces the five-state model, which includes new and terminated states. The transition from single-tasking (like DOS) to multi-tasking and true parallelism with multiple processors is highlighted. Key concepts include: - **Time Slices**: The length of time a process runs before being scheduled again affects system performance. Longer slices improve throughput but can reduce responsiveness, while shorter slices enhance interactivity but may lead to excessive scheduling overhead. - **Process Types**: Processes are categorized as CPU-bound (requiring significant CPU resources) or I/O-bound (often waiting for I/O operations). Each type has different time slice requirements. - **Scheduler Types**: Cooperative schedulers allow processes to run until completion without intervention, while preemptive schedulers allocate time slices, suspending processes when their time is up. - **Completely Fair Scheduler (CFS)**: This non-time-slice scheduler allocates CPU time equally among processes, adjusting based on priority and weight, aiming for fairness in process execution. Overall, the document emphasizes the importance of effective scheduling strategies to optimize system performance and user experience.

Course Content#

Process Scheduling#

  • Process scheduling is a kernel subsystem that users cannot manage
  • The main task of process scheduling is to decide which ready state process runs first
  • A ready process is a non-blocking process that meets the running conditions and does not need to block and wait
  • A blocking process is a process that is sleeping [not using CPU] and needs to be awakened by the kernel

Three-State Model#

  • Blocking, Ready, Running
  • Image
  • Blocking does not occupy system resources; Ready needs to wait for kernel scheduling; Running blocks when there is an IO request or waiting for an event

👉 Five-State Model

  • Image
  • Adds New and Terminated states

From Single Task to Multi-Task#

  • DOS is a single-task operating system, running only one process at a time
    • Processes run in an interleaved manner, similar to multi-tasking
  • Multi-processor systems can achieve true parallelism
    • Concurrency VS Parallelism
    • Concurrency: Many tasks are in progress within a time period
    • Parallelism: The number of CPUs determines how many tasks can run simultaneously
  • The core of scheduling: which process to run and for how long

Time Slice#

Strongly influences the system's global behavior and performance

  • Too long
    • 👍: Increases system throughput and global performance
    • ×: Reduces concurrent execution; users feel noticeable latency [decreased responsiveness, too many applications can also lead to longer scheduling intervals]
  • Too short
    • 👍: Improves interactive performance
    • ×: A lot of time spent on scheduling; the performance improvement brought by temporal locality is greatly reduced
      • Temporal locality: Caching, memory [cache is built up over time]
      • Spatial locality: Preparing data from nearby spaces in advance
      • Implies learning and prediction
  • Therefore, the length of the time slice is difficult to determine; the solution is to simply not use time slices, replaced by a Completely Fair scheduling algorithm
    • The system adjusts based on the ready queue and priority strategy
    • The allocation of scheduling time and priority is essentially to better serve users and is not unfair
    • See below for details—Completely Fair Scheduler

Processor-Bound and IO-Bound Processes#

Divided based on the characteristics of CPU usage, depending on whether the processor or IO is used more

  • Processor-Bound
    • [Processes that consistently consume the available time slice]
    • Require a large amount of CPU resources and will consume all CPU allocated by the scheduler
    • For example: while(1); can easily reach 100% CPU usage
    • Requirements for time slice
      • Expect longer time slices to maximize cache hits and complete tasks quickly [will not waste resources]
  • IO-Bound
    • [Processes that spend most of their time in a blocking state or waiting for resources]
    • Often block on file IO operations: files, network, keyboard, mouse [like listening to music, typing]; or do nothing except request the kernel to perform I/O operations [like cp]
    • Requirements for time slice
      • Need shorter time slices; the quicker they are scheduled, the more seamless the experience feels
      • Processes will only run for a very short time and then block on kernel resources; will use interrupts

Scheduler Classification#

Cooperative Scheduler#

[Non-intervention type, low practicality]

  • Processes will run until they finish themselves
  • The operating system does not intervene at all

Preemptive Scheduler#

[Time slice type]

  • The kernel allocates a time slice to the process; when the time slice is used up, the process is suspended [not using any resources], and other processes are executed
    • If there are no other ready processes, the kernel reallocates a time slice to the already executed process
  • The scheduler arranges the order and duration of process execution
  • [PS]
    • New state enters the ready queue, terminated state exits the ready queue [Five-State Model]
    • Considers priority, all processes have a chance to run

Completely Fair Scheduler#

[Non-time slice type]

CFS: Completely Fair Scheduler

  • For N processes, each process is allocated 1 / N of the processor time
  • Priority and weight
    • The higher the priority, the smaller the value, the greater the weight
    • Default priority is 0, weight is 1
  • Target Latency: Fixed period for scheduling
  • Minimum Granularity: Avoids too small target latency leading to excessively short run times for each process

👉 Fair Quota

Tips#

Loading...
Ownership of this post data is guaranteed by blockchain and smart contracts to the creator alone.