首先,要实现并发, 同时运行的进程必须是没有先后顺序的,否则也不适用.比如对多台server进行ping.
一. 通过后台运行的方式
常规方式:
[root@vm57-120 ~]# cat ping.sh
#!/bin/bash
for ip in 192.168.57.121 192.168.57.122 192.168.57.123
do
{ ping -c 20 $ip >/tmp/$ip.log; }
done
wait
wc -l /tmp/*.log
# time sh ping.sh
25 /tmp/192.168.57.121.log
25 /tmp/192.168.57.122.log
25 /tmp/192.168.57.123.log
75 total
real 0m57.073s
user 0m0.012s
sys 0m0.014s
后台运行方式:
[root@vm57-120 ~]# cat mping.sh
#!/bin/bash
for ip in 192.168.57.121 192.168.57.122 192.168.57.123
do
{ ping -c 20 $ip >/tmp/$ip.log; } &
done
wait
wc -l /tmp/*.log
[root@vm57-120 ~]# time sh mping.sh
25 /tmp/192.168.57.121.log
25 /tmp/192.168.57.122.log
25 /tmp/192.168.57.123.log
75 total
real 0m19.037s
user 0m0.014s
sys 0m0.023s
以上两种运行方式在效率上的差距是很明显的. 另外上述的这种"并发"的进程数是不可控的,即你的循环上限是多少,并发的进程数也会无限接近这个值.
二. 通过管道实现进程数可控
[root@vm57-120 ~]# vim eping.sh
#!/bin/bash
pipefile=$$.fifo
mkfifo $pipefile
exec 3<>$pipefile
rm $pipefile
iplist=(192.168.57.121 192.168.57.122 192.168.57.123)
thread=5;
for (( i = 1;i<=${thread};i++ ))
do
echo;
done >&3; # &3代表引用文件描述符3,这里循环就是向这个描述符里插入了5个占位符
#
for ip in ${iplist[*]}
do
read -u 3; # 描述符是按行插入和读取的,每次读取一次将少一个占位符,直到全部都读取完后则会等待,等待之前的线程任务完成才可以继续新的任务,这样就达到了控制任务数的目的。
( ping -c 20 $ip >/tmp/$ip.log;echo >&3 ) &
done
wait
exec 3>&-; #关闭定义的管道
wc -l /tmp/*.log
[root@vm57-120 ~]# time sh eping.sh
25 /tmp/192.168.57.121.log
25 /tmp/192.168.57.122.log
25 /tmp/192.168.57.123.log
75 total
real 0m19.037s
user 0m0.014s
sys 0m0.030s
这与上面的时间是一样的,我们调整进程数thread=2:
[root@vm57-120 ~]# time sh eping.sh
25 /tmp/192.168.57.121.log
25 /tmp/192.168.57.122.log
25 /tmp/192.168.57.123.log
75 total
real 0m38.042s
user 0m0.012s
sys 0m0.032s
{ } &大括号丢到后台时,最后需要一个分号; { cmd1;cmd2;} &
( cmd ) & 小括号必须放到脚本里运行才可以;在当前shell运行有问题;
阅读(1421) | 评论(0) | 转发(0) |