preface:

    Above, we introduced the basic API usage of FileChannel. In this article, let's take a look at the high-level API in FileChannel.

     It's high-level. Really, these knowledge points make a lot of use of the advanced playing method of file transfer mapping of the operating system, which greatly improves the efficiency of our file operation. kafka and rocketMQ, which we are familiar with, also use these high-level API s to achieve such high efficiency.

     We propose a requirement, which is described as follows: provide an external socket service, which is to obtain the files in the specified file directory, write them into the socket, and finally display them on the client side.

1. Traditional file network transmission process

    According to this requirement, we use the following code to complete it in a conventional way:

File file = new File("D:\\test.txt");
Long size = file.length();
byte[] arr = new byte[size.intValue()];

try {
    // 1. Read the contents of test.txt file into arr
    FileInputStream fileInputStream = new FileInputStream(file);
    fileInputStream.read(arr);

    // 2. Provision of external services
    Socket socket = new ServerSocket(9999).accept();

    // 3. Transfer to client
    socket.getOutputStream().write(arr);
} catch (FileNotFoundException e) {
    e.printStackTrace();
} catch (IOException e) {
    e.printStackTrace();
}

The above is an implementation of the simplest version.

So from the perspective of the operating system, what are the above transmission processes?

2.mmap optimization

File file = new File("D:\\test.txt");
Long size = file.length();
byte[] arr = new byte[size.intValue()];

try {
    // 1. Read the contents of test.txt file into arr
    RandomAccessFile raFile = new RandomAccessFile(file, "rwd");
    FileChannel channel = raFile.getChannel();
    MappedByteBuffer mappedByteBuffer = channel.map(FileChannel.MapMode.READ_WRITE, 0, size);

    // 2. Provision of external services
    ServerSocketChannel serverSocketChannel = ServerSocketChannel.open();

    serverSocketChannel.socket().bind(new InetSocketAddress(9999));
    serverSocketChannel.configureBlocking(false);

    while(true){
        SocketChannel socketChannel =
            serverSocketChannel.accept();

        if(socketChannel != null){
            // 3. Transfer to client
            socketChannel.write(mappedByteBuffer);
        }
    }

} catch (FileNotFoundException e) {
    e.printStackTrace();
} catch (IOException e) {
    e.printStackTrace();
}

So from the perspective of the operating system, what are the above transmission processes?

3.sendFile optimization (Linux version 2.1)

The sendFile function is provided in Linux version 2.1. What are the optimizations of this function for this example?

That is, the data can be transferred directly from the kernel file buffer to the Socket buffer without going through the user state
 

FileChannel provides transferTo (and transferFrom) methods to implement the sendFile function

File file = new File("D:\\test.txt");
Long size = file.length();

try {
    // 1. Read the contents of test.txt file into arr
    RandomAccessFile raFile = new RandomAccessFile(file, "rwd");
    FileChannel channel = raFile.getChannel();

    // 2. Provision of external services
    ServerSocketChannel serverSocketChannel = ServerSocketChannel.open();

    serverSocketChannel.socket().bind(new InetSocketAddress(9999));
    serverSocketChannel.configureBlocking(false);

    while(true){
        SocketChannel socketChannel =
            serverSocketChannel.accept();

        if(socketChannel != null){
            // 3. Use the transferTo method to transfer the file data to the client
            channel.transferTo(0, size, socketChannel);
        }
    }
} catch (FileNotFoundException e) {
    e.printStackTrace();
} catch (IOException e) {
    e.printStackTrace();
}

The code in the same 2 only uses a different method when transferring the file content to the socket in the last step. In this example, the FileChannel.transferTo method is used to transfer data.

So from the perspective of the operating system, what are the above transmission processes?

 

Referring to the four operation processes in 1, there is also less user space participation, and there is no switching between user state and kernel state.

So to sum up, there are two data copies and two context switches (compared with 3, the copy from the kernel file buffer to the kernel socket buffer is reduced)

Summary:

    Let's show the similarities and differences of the above four transmission modes through a chart