我喜欢解析行星（osm） planet-200309.xml 约1.26 TB）
为了计算使用java最少需要多长时间，我创建了一个小型sax解析应用程序：

final File f = new File("f:/planet-200309.xml");
    SAXParserFactory newInstance = SAXParserFactory.newInstance();
    final long start = System.currentTimeMillis();
    final CountingInputStream cif = new CountingInputStream(new FileInputStream(f));
    Thread t = new Thread() {
        public void run() {
            try {
                while (cif.available() > 0) {
                    Thread.sleep(10000);
                    long stop = System.currentTimeMillis();
                    long seconds = (stop - start) / 1000;
                    long bytesRead = cif.getBytesRead();
                    float bytePerSecond = bytesRead / seconds;
                    int expectedSeconds = (int) (f.length() / bytePerSecond);
                    System.out.println("Expected minutes: " + expectedSeconds / 60 + ", bytes per second:" + (int) bytePerSecond + " (reat: " + bytesRead
                            + ", took: " + seconds + ")");
                }
            } catch (IOException e) {
                e.printStackTrace();
            } catch (InterruptedException e) {
                e.printStackTrace();
            }
        };
    };
    t.start();

    newInstance.newSAXParser().parse(cif, new DefaultHandler() {
        @Override
        public void startElement(String uri, String localName, String qName, Attributes attributes) throws SAXException {
            if (!qName.equals("changeset") & !qName.equals("tag")) {
                System.out.println(qName);
            }
        }
    });

    long stop = System.currentTimeMillis();
    long took = stop - start;
    System.out.println(took / 1000);

我决定每10秒计算一下需要多长时间。这是我的输出：

Expected minutes: 174, bytes per second:122728576 (reat: 1227285768, took: 10)
Expected minutes: 173, bytes per second:123918400 (reat: 2478367973, took: 20)
Expected minutes: 172, bytes per second:124213368 (reat: 3726401103, took: 30)
Expected minutes: 175, bytes per second:122289280 (reat: 4891571271, took: 40)
Expected minutes: 186, bytes per second:115111152 (reat: 5755557747, took: 50)
Expected minutes: 197, bytes per second:108455448 (reat: 6507327092, took: 60)
Expected minutes: 212, bytes per second:100975920 (reat: 7068314710, took: 70)
Expected minutes: 224, bytes per second:95568256 (reat: 7645460576, took: 80)
Expected minutes: 236, bytes per second:90757120 (reat: 8168140838, took: 90)
Expected minutes: 237, bytes per second:90276424 (reat: 9027642303, took: 100)
Expected minutes: 240, bytes per second:89072968 (reat: 9798026674, took: 110)
Expected minutes: 250, bytes per second:85678456 (reat: 10281415054, took: 120)
Expected minutes: 257, bytes per second:83305712 (reat: 10829743052, took: 130)
Expected minutes: 256, bytes per second:83664016 (reat: 11712962690, took: 140)
Expected minutes: 250, bytes per second:85531576 (reat: 12829736785, took: 150)

在最初的10秒钟里，我计算了大约3个小时的时间消耗。两分钟后，我计算出4小时的时间消耗。
相互关联的事情是：似乎有一种趋势会减慢sax解析的速度。
我把这些值放到一个图表中：

这里有什么问题？为什么趋势会放缓？