Scalable IO in Java

Scalable IO in Java

http://gee.cs.oswego.edu/dl/cpjslides/nio.pdf

大部分IO都是下面这个步骤,

Most have same basic structure: 
Read request 
Decode request 
Process service 
Encode reply 
Send reply

关键是如何处理并发, 最原始就是单纯的用多线程

class Server implements Runnable {
    public void run() {
        try {
            ServerSocket ss = new ServerSocket(PORT);
            while (!Thread.interrupted())
            new Thread(new Handler(ss.accept())).start(); //创建新线程来handle
            // or, single-threaded, or a thread pool
        } catch (IOException ex) { /* ... */ }
    }

    static class Handler implements Runnable {
        final Socket socket;
        Handler(Socket s) { socket = s; }
        public void run() {
            try {
                byte[] input = new byte[MAX_INPUT];
                socket.getInputStream().read(input);
                byte[] output = process(input);
                socket.getOutputStream().write(output);
            } catch (IOException ex) { /* ... */ }
        }
        private byte[] process(byte[] cmd) { /* ... */ }
    }
}

显然简单的多线程会带来扩展性问题, 当client数量变的很多的时候, 还其他的可用性, 性能的问题 
解决方法就是Divide-and-conquer, 分开后, 就需要Event-driven Designs来串联起来...

单线程版本的Reactor, 所有事情read, process, write都由单个线程完成, 完成一步重新设置下一步的event, 然后干其他的事 
问题当然就是, 其中任何步骤不能消耗太多时间, 因为只有一个线程, 你占住了就会block其他任务 
ps, 不明白为什么要画那么大个acceptor, 只是作为第一步的callback对象...

看代码会更清楚,

class Reactor implements Runnable {
    final Selector selector;
    final ServerSocketChannel serverSocket;
    Reactor(int port) throws IOException { //Reactor初始化
        selector = Selector.open();
        serverSocket = ServerSocketChannel.open();
        serverSocket.socket().bind(new InetSocketAddress(port));
        serverSocket.configureBlocking(false); //非阻塞
        SelectionKey sk = serverSocket.register(selector, SelectionKey.OP_ACCEPT); //分步处理,第一步,接收accept事件
        sk.attach(new Acceptor()); //attach callback object, Acceptor
    }

    public void run() {
        try {
            while (!Thread.interrupted()) {
                selector.select();
                Set selected = selector.selectedKeys();
                Iterator it = selected.iterator();
                while (it.hasNext())
                    dispatch((SelectionKey)(it.next()); //Reactor负责dispatch收到的事件
                selected.clear();
            }
        } catch (IOException ex) { /* ... */ }
    }

    void dispatch(SelectionKey k) {
    	Runnable r = (Runnable)(k.attachment()); //调用之前注册的callback对象
    	if (r != null)
    	    r.run();
    }

    class Acceptor implements Runnable { // inner
        public void run() {
            try {
                SocketChannel c = serverSocket.accept();
                if (c != null)
                new Handler(selector, c);
            }
            catch(IOException ex) { /* ... */ }
        }
    }
}

final class Handler implements Runnable {
    final SocketChannel socket;
    final SelectionKey sk;
    ByteBuffer input = ByteBuffer.allocate(MAXIN);
    ByteBuffer output = ByteBuffer.allocate(MAXOUT);
    static final int READING = 0, SENDING = 1;
    int state = READING;

    Handler(Selector sel, SocketChannel c) throws IOException {
        socket = c; c.configureBlocking(false);
        // Optionally try first read now
        sk = socket.register(sel, 0);
        sk.attach(this); //将Handler作为callback对象
        sk.interestOps(SelectionKey.OP_READ); //第二步,接收Read事件
        sel.wakeup();
    }
    boolean inputIsComplete() { /* ... */ }
    boolean outputIsComplete() { /* ... */ }
    void process() { /* ... */ }

    public void run() {
        try {
            if (state == READING) read();
            else if (state == SENDING) send();
        } catch (IOException ex) { /* ... */ }
    }

    void read() throws IOException {
        socket.read(input);
        if (inputIsComplete()) {
            process();
            state = SENDING;
            // Normally also do first write now
            sk.interestOps(SelectionKey.OP_WRITE); //第三步,接收write事件
        }
    }
    void send() throws IOException {
        socket.write(output);
        if (outputIsComplete()) sk.cancel(); //write完就结束了, 关闭select key
    }
}

//上面 的实现用Handler来同时处理Read和Write事件, 所以里面出现状态判断
//我们可以用State-Object pattern来更优雅的实现
class Handler { // ...
    public void run() { // initial state is reader
        socket.read(input);
        if (inputIsComplete()) {
            process();
            sk.attach(new Sender());  //状态迁移, Read后变成write, 用Sender作为新的callback对象
              sk.interest(SelectionKey.OP_WRITE);
            sk.selector().wakeup();
        }
    }
    class Sender implements Runnable {
        public void run(){ // ...
            socket.write(output);
            if (outputIsComplete()) sk.cancel();
        }
    }
}

单线程模式的局限还是比较明显的 
所以改进是, 将比较耗时的部分, 从reactor线程中分离出去, 让reactor专门负责IO 
而另外创建Thread Pool和queue来缓存和处理任务 
所以其实已经进化成Proactor模式, 异步模式

class Handler implements Runnable {
    // uses util.concurrent thread pool
    static PooledExecutor pool = new PooledExecutor(...);
    static final int PROCESSING = 3;
    // ...
    synchronized void read() { // ...
        socket.read(input);
        if (inputIsComplete()) {
            state = PROCESSING;
            pool.execute(new Processer()); //使用线程pool异步执行
        }
    }

    synchronized void processAndHandOff() {
        process();
        state = SENDING; // or rebind attachment
        sk.interest(SelectionKey.OP_WRITE); //process完,开始等待write事件
    }

    class Processer implements Runnable {
        public void run() { processAndHandOff(); }
    }
}

使用多个reactor进程, 主reactor只负责accept, 然后将接收到的socketchannel交给subReactor去listen和处理 
当然也可以在subReactor下加上线程池进行异步处理 
坦白的说, 没看出用多个reactor有啥大的提升, 降低mainReactor listen的负担?

Selector[] selectors; //subReactors集合, 一个selector代表一个subReactor
int next = 0;
class Acceptor { // ...
    public synchronized void run() { ...
        Socket connection = serverSocket.accept(); //主selector负责accept
        if (connection != null)
            new Handler(selectors[next], connection); //选个subReactor去负责接收到的connection
        if (++next == selectors.length) next = 0;
    }
}

本文章摘自博客园，原文发布日期：2013-10-11