Asynchronous File I/O

Synchronous I/O means that the method is blocked until the I/O operation is complete, and then the method returns its data. With asynchronous I/O, a user can call BeginRead or BeginWrite. The main thread can continue doing other work, and later the user will be able to process the data. Also, multiple I/O requests can be pending simultaneously.

To be informed when this data is available, you can call EndRead or EndWrite, passing in the IAsyncResult corresponding to the I/O request you issued. You can also provide a callback method that should call EndRead or EndWrite to figure out how many bytes were read or written. Asynchronous I/O can offer better performance when many I/O requests are pending simultaneously, but generally requires some significant restructuring of your application to work correctly.

The Stream class supports the mixing of synchronous and asynchronous reads and writes on the same stream, regardless of whether the operating system allows this. Stream provides default implementations of asynchronous read and write operations in terms of their synchronous implementations, and provides default implementations of synchronous read and write operations in terms of their asynchronous implementations.

When implementing a class derived from Stream, it is necessary to provide an implementation for either the synchronous or the asynchronous Read and Write methods. While overriding Read and Write is allowable, and the default implementations of the asynchronous methods (BeginRead, EndRead, BeginWrite, and EndWrite) will work with your implementation of the synchronous methods, this does not provide the most efficient performance. Similarly, the synchronous Read and Write methods will work correctly if you provide an implementation of the asynchronous methods, but performance is generally better if you specifically implement the synchronous methods. The default implementations of ReadByte and WriteByte call the synchronous Read and Write methods with a one-element byte array. When deriving classes from Stream, if you have an internal byte buffer, it is strongly recommended that you override these methods to access your internal buffer for better performance.

A stream that connects to a backing store overrides either the synchronous or asynchronous Read and Write methods to get the functionality of the other by default. If a stream does not support asynchronous or synchronous operations, the implementer need only make the appropriate methods throw exceptions.

The following example is an asynchronous implementation of a hypothetical bulk image processor, followed by a synchronous implementation example. This code is designed to perform a CPU-intensive operation on every file in a directory. For more information, see the .NET Async Programming Model topics in the .NET Framework Developer Specifications.

using System;
using System.IO;
using System.Threading;
using System.Runtime.InteropServices;
using System.Runtime.Remoting.Messaging;

public class BulkImageProcAsync
{
    public const String ImageBaseName = "tmpImage-";
    public const int numImages = 200;
    public const int numPixels = 512*512;

    // ProcessImage has a simple O(N) loop, and you can vary the number
    // of times you repeat that loop to make the application more CPU-
    // bound or more IO-bound.
    public static int processImageRepeats = 20;

    // Threads must decrement NumImagesToFinish, and protect
    // their access to it through a mutex.
    public static int NumImagesToFinish = numImages;
    public static Object NumImagesMutex = new Object[0];
    // WaitObject is signalled when all image processing is done.
    public static Object WaitObject = new Object[0];
    public class ImageStateObject
    {
        public byte[] pixels;
        public int imageNum;
        public FileStream fs;
    }

    public static void MakeImageFiles()
    {
        int sides = (int) Math.Sqrt(numPixels);
        Console.Write("Making "+numImages+" "+sides+"x"+sides
            +" images...  ");
        byte[] pixels = new byte[numPixels];
        for(int i=0; i<numPixels; i++)
            pixels[i] = (byte) i;

        for(int i=0; i<numImages; i++) 
        {
            FileStream fs = new FileStream(ImageBaseName+i+".tmp", 
                FileMode.Create, FileAccess.Write, FileShare.None, 
                    8192, false);
            fs.Write(pixels, 0, pixels.Length);
            FlushFileBuffers(fs.Handle);
            fs.Close();
        }
        Console.WriteLine("Done.");
    }

    public static void ReadInImageCallback(IAsyncResult asyncResult)
    {
        ImageStateObject state = (ImageStateObject) asyncResult.AsyncState;
        Stream stream = state.fs;
        int bytesRead = stream.EndRead(asyncResult);
        if (bytesRead != numPixels)
            throw new Exception("In ReadInImageCallback, got the wrong 
                number of bytes from the image: {0}.", bytesRead);
        ProcessImage(state.pixels, state.imageNum);
        stream.Close();

        // Now write out the image.  
        // Using asynchronous I/O here appears not to be best practice.
        // It ends up swamping the threadpool, because the threadpool
        // threads are blocked on I/O requests that were just queued to
        // the threadpool. 
        FileStream fs = new 
            FileStream(ImageBaseName+state.imageNum+".done",
                FileMode.Create, FileAccess.Write, FileShare.None, 4096, 
                    false);
        fs.Write(state.pixels, 0, numPixels);
        fs.Close();

        // This application model uses too much memory.
        // Releasing memory as soon as possible is a good idea, 
        // especially global state.
        state.pixels = null;

        // Record that an image is finished now.
        lock(NumImagesMutex) 
        {
            NumImagesToFinish--;
            if (NumImagesToFinish==0) 
            {
                Monitor.Enter(WaitObject);
                Monitor.Pulse(WaitObject);
                Monitor.Exit(WaitObject);
            }
        }
    }

    public static void ProcessImage(byte[] pixels, int imageNum)
    {
        Console.WriteLine("ProcessImage "+imageNum);

        // Perform some CPU-intensive operation on the image.
        for(int i=0; i<processImageRepeats; i++)
            for(int j=0; j<numPixels; j++)
                pixels[j] += 1;
        Console.WriteLine("ProcessImage "+imageNum+" done.");
    }

    public static void ProcessImagesInBulk()
    {
        Console.WriteLine("Processing images...  ");
        long t0 = Environment.TickCount;
        NumImagesToFinish = numImages;
        AsyncCallback readImageCallback = new 
            AsyncCallback(ReadInImageCallback);
        for(int i=0; i<numImages; i++) 
        {
            ImageStateObject state = new ImageStateObject();
            state.pixels = new byte[numPixels];
            state.imageNum = i;
            // Very large items are read only once, so you can make the 
            // buffer on the FileStream very small to save memory.
            FileStream fs = new FileStream(ImageBaseName+i+".tmp",
                FileMode.Open, FileAccess.Read, FileShare.Read, 1, true);
            state.fs = fs;
            fs.BeginRead(state.pixels, 0, numPixels, readImageCallback,
                state);
        }

        // Determine whether all images are done being processed.  
        // If not, block until all are finished.
        bool mustBlock = false;
        lock (NumImagesMutex) 
        {
            if (NumImagesToFinish > 0)
                mustBlock = true;
        }
        if (mustBlock) 
        {
            Console.WriteLine("All worker threads are queued. Blocking
                until they complete. numLeft: {0}.", NumImagesToFinish);
            Monitor.Enter(WaitObject);
            Monitor.Wait(WaitObject);
            Monitor.Exit(WaitObject);
        }

        long t1 = Environment.TickCount;
        Console.WriteLine("Total time processing images: {0} ms",
            (t1-t0));
    }

    public static void Cleanup()
    {
        for(int i=0; i<numImages; i++) 
        {
            File.Delete(ImageBaseName+i+".tmp");
            File.Delete(ImageBaseName+i+".done");
        }
    }

    public static void TryToClearDiskCache()
    {
        // Try to force all pending writes to disk, and clear the
        // disk cache of any data.
        byte[] bytes = new byte[100*(1<<20)];
        for(int i=0; i<bytes.Length; i++)
            bytes[i] = 0;
        bytes = null;
        GC.Collect();
        Thread.Sleep(2000);
    }

    public static void Main(String[] args)
    {
        Console.WriteLine("Bulk image processing sample application, 
            using async IO.");
        Console.WriteLine("Simulates applying a simple transformation to
            "+numImages+" \"images\"");
        Console.WriteLine("(Async FileStream & Threadpool benchmark)");
        Console.WriteLine("Warning - this test requires "+(numPixels *
            numImages * 2)+" bytes of temporary space");

        if (args.Length==1) 
        {
            processImageRepeats = Int32.Parse(args[0]);
            Console.WriteLine("ProcessImage inner loop - {0}.",
                processImageRepeats);
        }
        MakeImageFiles();
        TryToClearDiskCache();
        ProcessImagesInBulk();
        Cleanup();
    }
    [DllImport("KERNEL32", SetLastError=true)]
    private static extern void FlushFileBuffers(IntPtr handle);
}

Here is a synchronous example of the same idea.

using System;
using System.IO;
using System.Threading;
using System.Runtime.InteropServices;
using System.Runtime.Remoting.Messaging;

public class BulkImageProcSync
{
    public const String ImageBaseName = "tmpImage-";
    public const int numImages = 200;
    public const int numPixels = 512*512;

    // ProcessImage has a simple O(N) loop, and you can vary the number
    // of times you repeat that loop to make the application more CPU-
    // bound or more IO-bound.
    public static int processImageRepeats = 20;

    public static void MakeImageFiles()
    {
        int sides = (int) Math.Sqrt(numPixels);
        Console.Write("Making "+numImages+" "+sides+"x"+sides+" 
            images...  ");
        byte[] pixels = new byte[numPixels];
        for(int i=0; i<numPixels; i++)
            pixels[i] = (byte) i;

        for(int i=0; i<numImages; i++) 
        {
            FileStream fs = new FileStream(ImageBaseName+i+".tmp",
                FileMode.Create, FileAccess.Write, FileShare.None, 
                    8192, false);
            fs.Write(pixels, 0, pixels.Length);
            FlushFileBuffers(fs.Handle);
            fs.Close();
        }
        Console.WriteLine("Done.");
    }

    public static void ProcessImage(byte[] pixels, int imageNum)
    {
        Console.WriteLine("ProcessImage "+imageNum);

        // Perform some CPU-intensive operation on the image.
        for(int i=0; i<processImageRepeats; i++)
            for(int j=0; j<numPixels; j++)
                pixels[j] += 1;
        Console.WriteLine("ProcessImage "+imageNum+" done.");
    }

    public static void ProcessImagesInBulk()
    {
        Console.WriteLine("Processing images...  ");
        long t0 = Environment.TickCount;

        byte[] pixels = new byte[numPixels];
        for(int i=0; i<numImages; i++) 
        {
            FileStream input = new FileStream(ImageBaseName+i+".tmp",
                FileMode.Open, FileAccess.Read, FileShare.Read, 
                    4196, false);
            input.Read(pixels, 0, numPixels);
            input.Close();
            ProcessImage(pixels, i);
            FileStream output = new FileStream(ImageBaseName+i+".done", 
                FileMode.Create, FileAccess.Write, FileShare.None, 
                    4196, false);
            output.Write(pixels, 0, numPixels);
            output.Close();
        }

        long t1 = Environment.TickCount;
        Console.WriteLine("Total time processing images: {0} ms", 
            (t1-t0));
    }

    public static void Cleanup()
    {
        for(int i=0; i<numImages; i++) 
        {
            File.Delete(ImageBaseName+i+".tmp");
            File.Delete(ImageBaseName+i+".done");
        }
    }

    public static void TryToClearDiskCache()
    {
        byte[] bytes = new byte[100*(1<<20)];
        for(int i=0; i<bytes.Length; i++)
            bytes[i] = 0;
        bytes = null;
        GC.Collect();
        Thread.Sleep(2000);
    }

    public static void Main(String[] args)
    {
        Console.WriteLine("Bulk image processing sample application, 
            using synchronous I/O");
        Console.WriteLine("Simulates applying a simple transformation to 
            "+numImages+" \"images\"");
        Console.WriteLine("(ie, Sync FileStream benchmark)");
        Console.WriteLine("Warning - this test requires "+(numPixels * 
            numImages * 2)+" bytes of tmp space");

        if (args.Length==1) 
        {
            processImageRepeats = Int32.Parse(args[0]);
            Console.WriteLine("ProcessImage inner loop – 
                "+processImageRepeats);
        }

        MakeImageFiles();
        TryToClearDiskCache();
        ProcessImagesInBulk();
        Cleanup();
    }

    [DllImport("KERNEL32", SetLastError=true)]
    private static extern void FlushFileBuffers(IntPtr handle);
}

See Also

Working with I/O | Stream | Stream.Read | Stream.Write | Stream.BeginRead | Stream.BeginWrite | Stream.EndRead | Stream.EndWrite | IAsyncResult | Mutex