下面让我们在请求-应答模式中编写一个小型的消息队列代理装置。
在Hello World客户/服务模型中，一个客户端和一个服务端进行通信。但在真实环境中，我们会需要让多个客户端和多个服务端进行通信。关键问题在于，服务端应该是无状态的，所有的状态都应该包含在一次请求中，或者存放其它介质中，如数据库。
我们有两种方式来连接多个客户端和多个服务端。第一种是让客户端直接和多个服务端进行连接。客户端套接字可以连接至多个服务端套接字，它所发送的请求会通过负载均衡的方式分发给服务端。比如说，有一个客户端连接了三个服务端，A、B、C，客户端产生了R1、R2、R3、R4四个请求，那么，R1和R4会由服务A处理，R2由B处理，R3由C处理：
这种设计的好处在于可以方便地添加客户端，但若要添加服务端，那就得修改每个客户端的配置。如果你有100个客户端，需要添加三个服务端，那么这些客户端都需要重新进行配置，让其知道新服务端的存在。
这种方式肯定不是我们想要的。一个网络结构中如果有太多固化的模块就越不容易扩展。因此，我们需要有一个模块位于客户端和服务端之间，将所有的知识都汇聚到这个网络拓扑结构中。理想状态下，我们可以任意地增减客户端或是服务端，不需要更改任何组件的配置。
下面就让我们编写这样一个组件。这个代理会绑定到两个端点，前端端点供客户端连接，后端端点供服务端连接。它会使用zmq_poll()来轮询这两个套接字，接收消息并进行转发。装置中不会有队列的存在，因为ZMQ已经自动在套接字中完成了。
在使用REQ和REP套接字时，其请求-应答的会话是严格同步。客户端发送请求，服务端接收请求并发送应答，由客户端接收。如果客户端或服务端中的一个发生问题（如连续两次发送请求），程序就会报错。
但是，我们的代理装置必须要是非阻塞式的，虽然可以使用zmq_poll()同时处理两个套接字，但这里显然不能使用REP和REQ套接字。
幸运的是，我们有DEALER和ROUTER套接字可以胜任这项工作，进行非阻塞的消息收发。DEALER过去被称为XREQ，ROUTER被称为XREP，但新的代码中应尽量使用DEALER/ROUTER这种名称。在第三章中你会看到如何用DEALER和ROUTER套接字构建不同类型的请求-应答模式。
下面就让我们看看DEALER和ROUTER套接字是怎样在装置中工作的。
下方的简图描述了一个请求-应答模式，REQ和ROUTER通信，DEALER再和REP通信。ROUTER和DEALER之间我们则需要进行消息转发：
请求-应答代理会将两个套接字分别绑定到前端和后端，供客户端和服务端套接字连接。在使用该装置之前，还需要对客户端和服务端的代码进行调整。
Client:
#include "../zhelpers.h" int main (void) { void *context = zmq_ctx_new (); // Socket to talk to server void *requester = zmq_socket (context, ZMQ_REQ); zmq_connect (requester, "tcp://localhost:5559"); int request_nbr; for (request_nbr = 0; request_nbr != 10; request_nbr++) { s_send (requester, "Hello"); char *string = s_recv (requester); printf ("Received reply %d [%s] ", request_nbr, string); free (string); } zmq_close (requester); zmq_ctx_destroy (context); return 0; }
Server:
#include "../zhelpers.h" #include <unistd.h> int main (void) { void *context = zmq_ctx_new (); // Socket to talk to clients void *responder = zmq_socket (context, ZMQ_REP); zmq_connect (responder, "tcp://localhost:5560"); while (1) { // Wait for next request from client char *string = s_recv (responder); printf ("Received request: [%s] ", string); free (string); // Do some 'work' sleep (1); // Send reply back to client s_send (responder, "World"); } // We never get here, but clean up anyhow zmq_close (responder); zmq_ctx_destroy (context); return 0; }
Broker
#include "../zhelpers.h" int main (void) { // Prepare our context and sockets void *context = zmq_ctx_new (); void *frontend = zmq_socket (context, ZMQ_ROUTER); void *backend = zmq_socket (context, ZMQ_DEALER); zmq_bind (frontend, "tcp://*:5559"); zmq_bind (backend, "tcp://*:5560"); // Initialize poll set zmq_pollitem_t items [] = { { frontend, 0, ZMQ_POLLIN, 0 }, { backend, 0, ZMQ_POLLIN, 0 } }; // Switch messages between sockets while (1) { zmq_msg_t message; zmq_poll (items, 2, -1); if (items [0].revents & ZMQ_POLLIN) { while (1) { // Process all parts of the message zmq_msg_init (&message); zmq_msg_recv (&message, frontend, 0); int more = zmq_msg_more (&message); zmq_msg_send (&message, backend, more? ZMQ_SNDMORE: 0); zmq_msg_close (&message); if (!more) break; // Last message part } } if (items [1].revents & ZMQ_POLLIN) { while (1) { // Process all parts of the message zmq_msg_init (&message); zmq_msg_recv (&message, backend, 0); int more = zmq_msg_more (&message); zmq_msg_send (&message, frontend, more? ZMQ_SNDMORE: 0); zmq_msg_close (&message); if (!more) break; // Last message part } } } // We never get here, but clean up anyhow zmq_close (frontend); zmq_close (backend); zmq_ctx_destroy (context); return 0; }
out:
// client Received reply 0 [World] Received reply 1 [World] Received reply 2 [World] Received reply 3 [World] Received reply 4 [World] Received reply 5 [World] Received reply 6 [World] Received reply 7 [World] Received reply 8 [World] Received reply 9 [World] // server Received request: [Hello] Received request: [Hello] Received request: [Hello] Received request: [Hello] Received request: [Hello] Received request: [Hello] Received request: [Hello] Received request: [Hello] Received request: [Hello] Received request: [Hello]
使用请求-应答代理可以让你的C/S网络结构更易于扩展：客户端不知道服务端的存在，服务端不知道客户端的存在。网络中唯一稳定的组件是中间的代理装置：
这里运行多个客户端看一下效果，同时改一下客户端服务端代码：
// client for (request_nbr = 0; request_nbr != 10; request_nbr++) { char strDst[256] = {0}; snprintf(strDst,256,"Hello %d",request_nbr); s_send (requester, strDst); char *string = s_recv (requester); printf ("Received reply %d [%s] ", request_nbr, string); free (string); } // server while (1) { // Wait for next request from client char *string = s_recv (responder); printf ("Received request: [%s] ", string); // Do some 'work' sleep (1); // Send reply back to client s_send (responder, string); free (string); }
out：
// client 1 Received reply 0 [Hello 0] Received reply 1 [Hello 1] Received reply 2 [Hello 2] Received reply 3 [Hello 3] Received reply 4 [Hello 4] Received reply 5 [Hello 5] Received reply 6 [Hello 6] Received reply 7 [Hello 7] Received reply 8 [Hello 8] Received reply 9 [Hello 9] // client 2 Received reply 0 [Hello 0] Received reply 1 [Hello 1] Received reply 2 [Hello 2] Received reply 3 [Hello 3] Received reply 4 [Hello 4] Received reply 5 [Hello 5] Received reply 6 [Hello 6] Received reply 7 [Hello 7] Received reply 8 [Hello 8] Received reply 9 [Hello 9] // server Received request: [Hello 0] Received request: [Hello 1] Received request: [Hello 2] Received request: [Hello 0] Received request: [Hello 3] Received request: [Hello 1] Received request: [Hello 4] Received request: [Hello 2] Received request: [Hello 5] Received request: [Hello 3] Received request: [Hello 6] Received request: [Hello 4] Received request: [Hello 7] Received request: [Hello 5] Received request: [Hello 8] Received request: [Hello 6] Received request: [Hello 9] Received request: [Hello 7] Received request: [Hello 8] Received request: [Hello 9]
可以看到客户端可以工作的很好，zmq帮我们正确的代理了消息。