Question

我正在使用HTTP Range请求部分下载文件并将它们合并在一起。这适用于HTTP，但使用HTTPS时，inputStream读取的字节数非常少，需要很长时间才能完成。此外，我不想删除Tread.Sleep(100)因为CPU使用率上升到100％。

当前源代码：

@Override
public Boolean call() throws Exception
{
    //Preparing Connection and setting ranges.
    HttpURLConnection urlConnection =
            (HttpURLConnection) fileLocation.openConnection();
    urlConnection.setRequestProperty(
    "Range", "Bytes=" + chunkStart + "-" + chunkEnd);

    InputStream inputStream = urlConnection.getInputStream();
    if (urlConnection.getResponseCode() != HttpURLConnection.HTTP_PARTIAL)
    {
        throw new Exception("Server returned status Code: " +
        urlConnection.getResponseCode());
    }

    ByteBuffer buffer = ByteBuffer.allocate((int)(chunkEnd - chunkStart));
    long bytesRead = 0;
    // Copying the inputStream to my byteBuffer
    ReadableByteChannel inputChannel = Channels.newChannel(inputStream);
    while ((bytesRead = inputChannel.read(buffer)) > 0)
    {
        System.out.println("Bytes read: " + bytesRead);
        Thread.sleep(100);
    }

    long writeOffset = chunkStart;
    buffer.flip();
    while (buffer.hasRemaining())
    {
      writeOffset += file.write(buffer, writeOffset);
    }
    buffer.clear();
  return true;
}

下载10mb文件。使用HTTP输出:(大约需要2秒）

 Bytes read: 197102
 Bytes read: 365844
 Bytes read: 218372
 Bytes read: 253512
 Bytes read: 336256
 ...

使用HTTPS输出:(大约需要5分钟）

Bytes read: 320
Bytes read: 1389
Bytes read: 1389
Bytes read: 1389
Bytes read: 1389
Bytes read: 1389
...

删除Thread.sleep会使用HTTPS提供此输出:(大约需要2秒，但CPU使用率约为99％）

Bytes Read: 1389
Bytes Read: 1389
Bytes Read: 1389
Bytes Read: 1389
Bytes Read: 1389
...
Bytes Read: 16384
Bytes Read: 16384
Bytes Read: 16384
Bytes Read: 16384

我还尝试将inputStream手动读取到byteArray，然后将其写入byteBuffer或使用BufferedInputStream，但我得到相同的结果。

我在这个主题上发现了一个非常相似的案例，但我不确定是否这样真的是正确的解决方案Java ssl performance issue in connection with downloads

更新：我查看了System.setProperty("javax.net.debug","ssl:packet")的SSLSocket调试信息，这就是我得到的

TLSv1.2 : TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256
BufferSize: 16916
Packet Buffer Size: 16921
[Raw read]: length = 1413

0000: 00 00 00 00 00 00 00 01   CD 87 B9 AD B3 60 94 93  .............`..
0010: D5 13 0E 5A 3F D0 5D CE   C1 46 93 62 23 67 77 57  ...Z?.]..F.b#gwW
0020: 39 D4 07 46 96 89 72 12   2F 91 74 92 D8 C4 8F A4  9..F..r./.t.....
...more and more
0570: 04 E1 83 F6 7E 72 8A 38   5F EA 2B A0 F5 75 C3 57  .....r.8_.+..u.W
0580: 78 23 5E 01 AB                                     x#^..
pool-1-thread-1, READ: TLSv1.2 Application Data, length = 1413
Padded plaintext after DECRYPTION:  len = 1389

这就解释了为什么我每次调用只能读取1389个字节，但是如果所有缓冲区大小都大得多，并且如果我在没有Thread.sleep的情况下运行它，它将如何增加这个大小。它几乎被填满。

HTTP / HTTPS范围请求读取InputStream

0 个答案: