Linux System Programming 学习笔记(六) 进程调度

浏览数：26 / 时间：2015年06月20日

1. 进程调度

the process scheduler is the component of a kernel that selects which process to run next.

进程调度器需要使处理器使用率最大化，并且提供使多个进程并发执行的虚拟

Deciding which processes run, when, and for how long is the process scheduler‘s fundamental responsibility.

时间片：the scheduler allocates the process a “slice” of the processor‘s time

Linux进程调度算法CFS（ Completely Fair Scheduler）：enforce fair access to a resource among contending consumers

2. 时间片

如果时间片过大，那么挂起进程开始执行前的等待时间过长，这将减小并发执行的粒度，甚至用户觉察到延迟

如果时间片过小，那么系统在进程之间切换的时间花销将很大，时间局部性的优势将丧失

CPU密集型：processes are hungry for CPU time，比如科学计算、数学计算、图像处理

I/O密集型：spend more time blocked waiting for some resource than executing，often issuing and waiting for file or network I/O, blocking on keyboard input

3. CFS调度

传统Unix进程调度中，有两个最基础的变量：优先级和时间片

CFS引进 fair scheduling 算法，CFS赋予每个进程一定比例的处理器时间，而不是时间片

开始时，CFS赋予N个进程等同的 1/N 处理器时间，然后权衡每个进程比例进行动态调整

Processes with the default nice value of zero have a weight of one, so their proportion is unchanged

Processes with a smaller nice value (higher priority) receive a larger weight, increasing their fraction of the processor

process with a larger nice value (lower priority) receive a smaller weight, decreasing their fraction of the processor

为了确定每个进程每次运行的实际时间，CFS needs to divide the proportions into a fixed period

例如：20毫秒，2个相同优先级的进程，则每个进程赋予相同的权重和处理器时间比例，即每个进程10毫秒

如果是5个进程，则每个进程分配4毫秒，如果是200个进程呢？

按照常理应该是每个进程分配100微秒，但是由于上下文切换开销巨大，并且局部性优势丧失，这将严重影响系统的吞吐率

这时CFS引入另一个变量：minimum granularity (最小粒度)

minimum granularity是任一进程运行时间长度的下限，这将保证上下文切换开销占系统总时间开销的比例不会过大

By assigning proportions of the processor and not fixed timeslices,CFS is able to enforce fairness: each process gets its fair share of the processor

4. nice val

processes are assigned priorities that affect how long they run，Unix has historically called these priorities nice values

Legal nice values range from ?20 to 19 inclusive, with a default value of 0， nice值越大，优先级越低，nice值越小，优先级越高

/* increments a process‘s nice value by inc and returns the newly updated value */
#include <unistd.h>

int nice (int inc);

Passing 0 for inc is an easy way to obtain the current nice value

nice函数错误时返回-1，但是新的nice值也可能就是-1，为了区分函数错误和返回nice新值为-1，则如下使用：

int ret;
errno = 0;
ret = nice (10);    /* increase our nice by 10 */
if (ret == ?1 && errno != 0)
        perror ("nice");
else
        printf ("nice value is now %d\n", ret);

获取/设置优先级：

#include <sys/time.h>
#include <sys/resource.h>
int getpriority (int which, int who);
int setpriority (int which, int who, int prio);

/* returns the current process’s priority */
int ret;
ret = getpriority (PRIO_PROCESS, 0);
printf ("nice value is %d\n", ret);

/* sets the priority of all processes in the current process group to 10 */
int ret;
ret = setpriority (PRIO_PGRP, 0, 10);
if (ret == ?1)
        perror ("setpriority");

5. 处理器关联

the process scheduler must decide which processes run on each CPU

如果一个进程在一个CPU核上被调度，the process scheduler should aim to schedule it on the same CPU in the future

因为进程从一个CPU核迁移到另一个CPU核的代价是巨大的（主要是缓存影响）

if a process moves to a new CPU and writes new data into memory, the data in the old CPU‘s cache can become stale

这样，进程调度器必须在进程迁移CPU花销和多个CPU负载均衡之间取得平衡

The Linux scheduler attempts to schedule the same processes on the same processors for as long as possible, migrating a process from one CPU

to another only in situations of extreme load imbalance. This allows the scheduler to minimize the cache effects of migration but still ensure that

all processors in a system are evenly loaded

6. 实时系统

硬实时和软实时

A hard real-time system requires absolute adherence to operational deadlines

A soft real-time system does not consider over‐running a deadline to be a critical failure

7. Linux实时调度策略

FIFO调度：A FIFO-classed process will continue running so long as no higher-priority process becomes runnable

时间片轮转：When an RR-classed process exhausts its timeslice, the scheduler moves it to the end of the list of processes at its priority

注意：这两种实时调度策略中，如果存在更高优先级的进程，那么低优先级进程将不会运行

int policy;
/* get our scheduling policy */
policy = sched_getscheduler (0);
switch (policy) {
case SCHED_OTHER:
        printf ("Policy is normal\n");
        break;
case SCHED_RR:
        printf ("Policy is round-robin\n");
        break;
case SCHED_FIFO:
        printf ("Policy is first-in, first-out\n");
        break;
case -1:
        perror ("sched_getscheduler");
        break;
default:
        fprintf (stderr, "Unknown policy!\n");
}

Linux implements a range of 1 to 99 inclusive for the two real-time scheduling policies

Linux provides two system calls for retrieving the range of valid priority values：

int min, max;
min = sched_get_priority_min (SCHED_RR);
if (min == ?1) {
        perror ("sched_get_priority_min");
        return 1;
}
max = sched_get_priority_max (SCHED_RR);
if (max == ?1) {
        perror ("sched_get_priority_max");
        return 1;
}
printf ("SCHED_RR priority range is %d - %d\n", min, max);