什么要缓存curl资源

在看公司的代码框架底层时，发现了一个问题，如下：

代码中调用接口时，使用的是curl，框架将curl资源以IP ：端口的形式缓存了下来，例如:

10.10.10.10:80     curl1

10.10.10.10:8080 curl2

有点不太理解原因，所以作了两个实验，过程如下：

实验一：两个curl请求相同域名下的不同接口

实验二：一个curl请求相同域名下的不同接口

//两个curl请求相同域名的不同接口
$ch = curl_init();
curl_setopt($ch, CURLOPT_URL, "http://www.baidu.com/");
curl_setopt($ch, CURLOPT_HEADER, 0);
$r = curl_exec($ch);

$ch1 = curl_init();
curl_setopt($ch1, CURLOPT_URL, "http://www.baidu.com/img/bdlogo.gif");
curl_setopt($ch1, CURLOPT_HEADER, 0);
$r1 = curl_exec($ch1);

用strace跟踪系统调用，然后过滤里面有用的信息，命令如下：

strace php cron_test.php 1 >b 2>a
egrep "(fcntl|socket|sendto|connect|recvfrom)" a

结果如下：

socket(PF_INET, SOCK_STREAM, IPPROTO_TCP) = 3
fcntl(3, F_GETFL)                       = 0x2 (flags O_RDWR)
fcntl(3, F_SETFL, O_RDWR|O_NONBLOCK)    = 0
connect(3, {sa_family=AF_INET, sin_port=htons(80), sin_addr=inet_addr("61.135.169.125")}, 16) = -1 EINPROGRESS (Operation now in progress)
sendto(3, "GET / HTTP/1.1\r\nHost: www.baidu."..., 52, MSG_NOSIGNAL, NULL, 0) = 52
recvfrom(3, "HTTP/1.1 200 OK\r\nDate: Fri, 15 N"..., 16384, MSG_NOSIGNAL, NULL, NULL) = 1990
recvfrom(3, "round-image:url(http://s1.bdstat"..., 16384, MSG_NOSIGNAL, NULL, NULL) = 2896
recvfrom(3, "www.baidu.com%2F\" target=\"_blank"..., 16384, MSG_NOSIGNAL, NULL, NULL) = 1448
recvfrom(3, "http://baike.baidu.com\">\347\231\276\347\247\221</"..., 16384, MSG_NOSIGNAL, NULL, NULL) = 1448
recvfrom(3, "sp;<a onmousedown=\"return ns_c({"..., 16384, MSG_NOSIGNAL, NULL, NULL) = 2896
recvfrom(3, "1&ie=utf-8&s=s'],'tj_baike': ['w"..., 16384, MSG_NOSIGNAL, NULL, NULL) = 1792
socket(PF_INET, SOCK_STREAM, IPPROTO_TCP) = 4
fcntl(4, F_GETFL)                       = 0x2 (flags O_RDWR)
fcntl(4, F_SETFL, O_RDWR|O_NONBLOCK)    = 0
connect(4, {sa_family=AF_INET, sin_port=htons(80), sin_addr=inet_addr("61.135.169.125")}, 16) = -1 EINPROGRESS (Operation now in progress)
sendto(4, "GET /img/bdlogo.gif HTTP/1.1\r\nHo"..., 66, MSG_NOSIGNAL, NULL, 0) = 66
recvfrom(4, "HTTP/1.1 200 OK\r\nDate: Fri, 15 N"..., 16384, MSG_NOSIGNAL, NULL, NULL) = 2080

//一个curl请求相同域名下的两个接口
$ch = curl_init();
curl_setopt($ch, CURLOPT_URL, "http://www.baidu.com/");
curl_setopt($ch, CURLOPT_HEADER, 0);
$r1 = curl_exec($ch);
curl_setopt($ch, CURLOPT_URL, "http://www.baidu.com/img/bdlogo.gif");
$r2 = curl_exec($ch);

用strace跟踪系统调用，然后过滤里面有用的信息，命令如下：

strace php cron_test.php 1 >b 2>a
egrep "(fcntl|socket|sendto|connect|recvfrom)" a

结果如下：

connect(3, {sa_family=AF_INET, sin_port=htons(80), sin_addr=inet_addr("61.135.169.105")}, 16) = -1 EINPROGRESS (Operation now in progress)
sendto(3, "GET / HTTP/1.1\r\nHost: www.baidu."..., 52, MSG_NOSIGNAL, NULL, 0) = 52
recvfrom(3, "HTTP/1.1 200 OK\r\nDate: Fri, 15 N"..., 16384, MSG_NOSIGNAL, NULL, NULL) = 618
sendto(3, "GET /img/bdlogo.gif HTTP/1.1\r\nHo"..., 66, MSG_NOSIGNAL, NULL, 0) = 66
recvfrom(3, "HTTP/1.1 200 OK\r\nDate: Fri, 15 N"..., 16384, MSG_NOSIGNAL, NULL, NULL) = 2080

以上代码中，红色部分为两个实验的区别

可以看出，请求两个相同域名的接口时，用两个curl时，建立了两次连接，而用一个curl时，共用了一次连接

缓存curl可以节省一次网络连接

这就是为什么要缓存curl资源的原因。