基于Java回顧之網(wǎng)絡(luò)通信的應(yīng)用分析

更新時(shí)間：2013年05月08日 09:25:30 作者：

在這篇文章里，我們主要討論如何使用Java實(shí)現(xiàn)網(wǎng)絡(luò)通信，包括TCP通信、UDP通信、多播以及NIO

TCP連接

TCP的基礎(chǔ)是Socket，在TCP連接中，我們會(huì)使用ServerSocket和Socket，當(dāng)客戶端和服務(wù)器建立連接以后，剩下的基本就是對(duì)I/O的控制了。

我們先來(lái)看一個(gè)簡(jiǎn)單的TCP通信，它分為客戶端和服務(wù)器端。

客戶端代碼如下：

簡(jiǎn)單的TCP客戶端 
 import java.net.*;
 import java.io.*;
 public class SimpleTcpClient {

     public static void main(String[] args) throws IOException
     {
         Socket socket = null;
         BufferedReader br = null;
         PrintWriter pw = null;
         BufferedReader brTemp = null;
         try
         {
             socket = new Socket(InetAddress.getLocalHost(), 5678);
             br = new BufferedReader(new InputStreamReader(socket.getInputStream()));
             pw = new PrintWriter(socket.getOutputStream());
             brTemp = new BufferedReader(new InputStreamReader(System.in));
             while(true)
             {
                 String line = brTemp.readLine();
                 pw.println(line);
                 pw.flush();
                 if (line.equals("end")) break;
                 System.out.println(br.readLine());
             }
         }
         catch(Exception ex)
         {
             System.err.println(ex.getMessage());
         }
         finally
         {
             if (socket != null) socket.close();
             if (br != null) br.close();
             if (brTemp != null) brTemp.close();
             if (pw != null) pw.close();
         }
     }
 }

服務(wù)器端代碼如下：

復(fù)制代碼代碼如下:

簡(jiǎn)單版本TCP服務(wù)器端
 import java.net.*;
 import java.io.*;
 public class SimpleTcpServer {

     public static void main(String[] args) throws IOException
     {
         ServerSocket server = null;
         Socket client = null;
         BufferedReader br = null;
         PrintWriter pw = null;
         try
         {
             server = new ServerSocket(5678);
             client = server.accept();
             br = new BufferedReader(new InputStreamReader(client.getInputStream()));
             pw = new PrintWriter(client.getOutputStream());
             while(true)
             {
                 String line = br.readLine();
                 pw.println("Response:" + line);
                 pw.flush();
                 if (line.equals("end")) break;
             }
         }
         catch(Exception ex)
         {
             System.err.println(ex.getMessage());
         }
         finally
         {
             if (server != null) server.close();
             if (client != null) client.close();
             if (br != null) br.close();
             if (pw != null) pw.close();
         }
     }
 }

這里的服務(wù)器的功能非常簡(jiǎn)單，它接收客戶端發(fā)來(lái)的消息，然后將消息“原封不動(dòng)”的返回給客戶端。當(dāng)客戶端發(fā)送“end”時(shí)，通信結(jié)束。

上面的代碼基本上勾勒了TCP通信過(guò)程中，客戶端和服務(wù)器端的主要框架，我們可以發(fā)現(xiàn)，上述的代碼中，服務(wù)器端在任何時(shí)刻，都只能處理來(lái)自客戶端的一個(gè)請(qǐng)求，它是串行處理的，不能并行，這和我們印象里的服務(wù)器處理方式不太相同，我們可以為服務(wù)器添加多線程，當(dāng)一個(gè)客戶端的請(qǐng)求進(jìn)入后，我們就創(chuàng)建一個(gè)線程，來(lái)處理對(duì)應(yīng)的請(qǐng)求。

改善后的服務(wù)器端代碼如下：

復(fù)制代碼代碼如下:

多線程版本的TCP服務(wù)器端
 import java.net.*;
 import java.io.*;
 public class SmartTcpServer {
     public static void main(String[] args) throws IOException
     {
         ServerSocket server = new ServerSocket(5678);
         while(true)
         {
             Socket client = server.accept();
             Thread thread = new ServerThread(client);
             thread.start();
         }
     }
 }

 class ServerThread extends Thread
 {
     private Socket socket = null;

     public ServerThread(Socket socket)
     {
         this.socket = socket;
     }

     public void run() {
         BufferedReader br = null;
         PrintWriter pw = null;
         try
         {
             br = new BufferedReader(new InputStreamReader(socket.getInputStream()));
             pw = new PrintWriter(socket.getOutputStream());
             while(true)
             {
                 String line = br.readLine();
                 pw.println("Response:" + line);
                 pw.flush();
                 if (line.equals("end")) break;
             }
         }
         catch(Exception ex)
         {
             System.err.println(ex.getMessage());
         }
         finally
         {
             if (socket != null)
                 try {
                     socket.close();
                 } catch (IOException e1) {
                     e1.printStackTrace();
                 }
             if (br != null)
                 try {
                     br.close();
                 } catch (IOException e) {
                     e.printStackTrace();
                 }
             if (pw != null) pw.close();
         }
     }
 }

修改后的服務(wù)器端，就可以同時(shí)處理來(lái)自客戶端的多個(gè)請(qǐng)求了。

在編程的過(guò)程中，我們會(huì)有“資源”的概念，例如數(shù)據(jù)庫(kù)連接就是一個(gè)典型的資源，為了提升性能，我們通常不會(huì)直接銷毀數(shù)據(jù)庫(kù)連接，而是使用數(shù)據(jù)庫(kù)連接池的方式來(lái)對(duì)多個(gè)數(shù)據(jù)庫(kù)連接進(jìn)行管理，已實(shí)現(xiàn)重用的目的。對(duì)于Socket連接來(lái)說(shuō)，它也是一種資源，當(dāng)我們的程序需要大量的Socket連接時(shí)，如果每個(gè)連接都需要重新建立，那么將會(huì)是一件非常沒(méi)有效率的做法。

和數(shù)據(jù)庫(kù)連接池類似，我們也可以設(shè)計(jì)TCP連接池，這里的思路是我們用一個(gè)數(shù)組來(lái)維持多個(gè)Socket連接，另外一個(gè)狀態(tài)數(shù)組來(lái)描述每個(gè)Socket連接是否正在使用，當(dāng)程序需要Socket連接時(shí)，我們遍歷狀態(tài)數(shù)組，取出第一個(gè)沒(méi)被使用的Socket連接，如果所有連接都在使用，拋出異常。這是一種很直觀簡(jiǎn)單的“調(diào)度策略”，在很多開(kāi)源或者商業(yè)的框架中（Apache/Tomcat），都會(huì)有類似的“資源池”。

TCP連接池的代碼如下：

復(fù)制代碼代碼如下:

一個(gè)簡(jiǎn)單的TCP連接池
 import java.net.*;
 import java.io.*;
 public class TcpConnectionPool {

     private InetAddress address = null;
     private int port;
     private Socket[] arrSockets = null;
     private boolean[] arrStatus = null;
     private int count;

     public TcpConnectionPool(InetAddress address, int port, int count)
     {
         this.address = address;
         this.port = port;
         this .count = count;
         arrSockets = new Socket[count];
         arrStatus = new boolean[count];

         init();
     }

     private void init()
     {
         try
         {
             for (int i = 0; i < count; i++)
             {
                 arrSockets[i] = new Socket(address.getHostAddress(), port);
                 arrStatus[i] = false;
             }
         }
         catch(Exception ex)
         {
             System.err.println(ex.getMessage());
         }
     }

     public Socket getConnection()
     {
         if (arrSockets == null) init();
         int i = 0;
         for(i = 0; i < count; i++)
         {
             if (arrStatus[i] == false) 
             {
                 arrStatus[i] = true;
                 break;
             }
         }
         if (i == count) throw new RuntimeException("have no connection availiable for now.");

         return arrSockets[i];
     }

     public void releaseConnection(Socket socket)
     {
         if (arrSockets == null) init();
         for (int i = 0; i < count; i++)
         {
             if (arrSockets[i] == socket)
             {
                 arrStatus[i] = false;
                 break;
             }
         }
     }

     public void reBuild()
     {
         init();
     }

     public void destory()
     {
         if (arrSockets == null) return;

         for(int i = 0; i < count; i++)
         {
             try
             {
                 arrSockets[i].close();
             }
             catch(Exception ex)
             {
                 System.err.println(ex.getMessage());
                 continue;
             }
         }
     }
 }

UDP連接

UDP是一種和TCP不同的連接方式，它通常應(yīng)用在對(duì)實(shí)時(shí)性要求很高，對(duì)準(zhǔn)確定要求不高的場(chǎng)合，例如在線視頻。UDP會(huì)有“丟包”的情況發(fā)生，在TCP中，如果Server沒(méi)有啟動(dòng)，Client發(fā)消息時(shí)，會(huì)報(bào)出異常，但對(duì)UDP來(lái)說(shuō)，不會(huì)產(chǎn)生任何異常。

UDP通信使用的兩個(gè)類時(shí)DatagramSocket和DatagramPacket，后者存放了通信的內(nèi)容。

下面是一個(gè)簡(jiǎn)單的UDP通信例子，同TCP一樣，也分為Client和Server兩部分，Client端代碼如下：

復(fù)制代碼代碼如下:

UDP通信客戶端
 import java.net.*;
 import java.io.*;
 public class UdpClient {

     public static void main(String[] args)
     {
         try
         {
             InetAddress host = InetAddress.getLocalHost();
             int port = 5678;
             BufferedReader br = new BufferedReader(new InputStreamReader(System.in));
             while(true)
             {
                 String line = br.readLine();
                 byte[] message = line.getBytes();
                 DatagramPacket packet = new DatagramPacket(message, message.length, host, port);
                 DatagramSocket socket = new DatagramSocket();
                 socket.send(packet);
                 socket.close();
                 if (line.equals("end")) break;
             }
             br.close();
         }
         catch(Exception ex)
         {
             System.err.println(ex.getMessage());
         }
     }
 }

Server端代碼如下：

復(fù)制代碼代碼如下:

UDP通信服務(wù)器端
 import java.net.*;
 import java.io.*;
 public class UdpServer {

     public static void main(String[] args)
     {
         try
         {
             int port = 5678;
             DatagramSocket dsSocket = new DatagramSocket(port);
             byte[] buffer = new byte[1024];
             DatagramPacket packet = new DatagramPacket(buffer, buffer.length);
             while(true)
             {
                 dsSocket.receive(packet);
                 String message = new String(buffer, 0, packet.getLength());
                 System.out.println(packet.getAddress().getHostName() + ":" + message);
                 if (message.equals("end")) break;
                 packet.setLength(buffer.length);
             }
             dsSocket.close();
         }
         catch(Exception ex)
         {
             System.err.println(ex.getMessage());
         }
     }
 }

這里，我們也假設(shè)和TCP一樣，當(dāng)Client發(fā)出“end”消息時(shí)，認(rèn)為通信結(jié)束，但其實(shí)這樣的設(shè)計(jì)不是必要的，Client端可以隨時(shí)斷開(kāi)，并不需要關(guān)心Server端狀態(tài)。
多播（Multicast）

多播采用和UDP類似的方式，它會(huì)使用D類IP地址和標(biāo)準(zhǔn)的UDP端口號(hào)，D類IP地址是指224.0.0.0到239.255.255.255之間的地址，不包括224.0.0.0。

多播會(huì)使用到的類是MulticastSocket，它有兩個(gè)方法需要關(guān)注：joinGroup和leaveGroup。

下面是一個(gè)多播的例子，Client端代碼如下：

復(fù)制代碼代碼如下:

多播通信客戶端
 import java.net.*;
 import java.io.*;
 public class MulticastClient {

     public static void main(String[] args)
     {
         BufferedReader br = new BufferedReader(new InputStreamReader(System.in));
         try
         {
             InetAddress address = InetAddress.getByName("230.0.0.1");
             int port = 5678;
             while(true)
             {
                 String line = br.readLine();
                 byte[] message = line.getBytes();
                 DatagramPacket packet = new DatagramPacket(message, message.length, address, port);
                 MulticastSocket multicastSocket = new MulticastSocket();
                 multicastSocket.send(packet);
                 if (line.equals("end")) break;
             }
             br.close();
         }
         catch(Exception ex)
         {
             System.err.println(ex.getMessage());
         }
     }
 }

服務(wù)器端代碼如下：

復(fù)制代碼代碼如下:

多播通信服務(wù)器端
 import java.net.*;
 import java.io.*;
 public class MulticastServer {

     public static void main(String[] args)
     {
         int port = 5678;
         try
         {
             MulticastSocket multicastSocket = new MulticastSocket(port);
             InetAddress address = InetAddress.getByName("230.0.0.1");
             multicastSocket.joinGroup(address);
             byte[] buffer = new byte[1024];
             DatagramPacket packet = new DatagramPacket(buffer, buffer.length);
             while(true)
             {
                 multicastSocket.receive(packet);
                 String message = new String(buffer, packet.getLength());
                 System.out.println(packet.getAddress().getHostName() + ":" + message);
                 if (message.equals("end")) break;
                 packet.setLength(buffer.length);
             }
             multicastSocket.close();
         }
         catch(Exception ex)
         {
             System.err.println(ex.getMessage());
         }
     }
 }

NIO（New IO）

NIO是JDK1.4引入的一套新的IO API，它在緩沖區(qū)管理、網(wǎng)絡(luò)通信、文件存取以及字符集操作方面有了新的設(shè)計(jì)。對(duì)于網(wǎng)絡(luò)通信來(lái)說(shuō)，NIO使用了緩沖區(qū)和通道的概念。

下面是一個(gè)NIO的例子，和我們上面提到的代碼風(fēng)格有很大的不同。

復(fù)制代碼代碼如下:

NIO例子
 import java.io.*;
 import java.nio.*;
 import java.nio.channels.*;
 import java.nio.charset.*;
 import java.net.*;
 public class NewIOSample {

     public static void main(String[] args)
     {
         String host="127.0.0.1";
         int port = 5678;
         SocketChannel channel = null;
         try
         {
             InetSocketAddress address = new InetSocketAddress(host,port);
             Charset charset = Charset.forName("UTF-8");
             CharsetDecoder decoder = charset.newDecoder();
             CharsetEncoder encoder = charset.newEncoder();

             ByteBuffer buffer = ByteBuffer.allocate(1024);
             CharBuffer charBuffer = CharBuffer.allocate(1024);

             channel = SocketChannel.open();
             channel.connect(address);

             String request = "GET / \r\n\r\n";
             channel.write(encoder.encode(CharBuffer.wrap(request)));

             while((channel.read(buffer)) != -1)
             {
                 buffer.flip();
                 decoder.decode(buffer, charBuffer, false);
                 charBuffer.flip();
                 System.out.println(charBuffer);
                 buffer.clear();
                 charBuffer.clear();
             }
         }
         catch(Exception ex)
         {
             System.err.println(ex.getMessage());
         }
         finally
         {
             if (channel != null)
                 try {
                     channel.close();
                 } catch (IOException e) {
                     // TODO Auto-generated catch block
                     e.printStackTrace();
                 }
         }
     }
 }