Java Apache FOP transformation.transform需要花费大量时间

我们使用基于模板的方法生成PDF文件，为此，我们使用velocity作为模板引擎，使用ApacheFop作为PDF引擎

我们需要生成可以包含0.2M表行的PDF。 下面是我们编写的代码。MyBean是JavaBean类，它将有0.2行

velocityContext.put("MyBean", myBean);
final ByteArrayOutputStream outStream = new ByteArrayOutputStream();
fileWriter writer1 = new FileWriter("C:\\test\\output.fo");

template.merge(velocityContext, writer1);
writer1.flush();
writer1.close();

我的速度模板是这样的

<?xml version="1.0" encoding="utf-8"?>
   <fo:root xmlns:fo="http://www.w3.org/1999/XSL/Format">
     <fo:layout-master-set>
        <fo:simple-page-master master-name="mybean"
                               page-height="29.7cm"
                               page-width="37cm"
                               margin-top="1cm"
                               margin-bottom="2cm"
                               margin-left="1cm"
                               margin-right="1cm">
            <fo:region-body margin-top="3cm"/>
            <fo:region-before extent="3cm"/>
            <fo:region-after extent="1.5cm"/>
        </fo:simple-page-master>
    </fo:layout-master-set>

  #foreach($grid in $MyBean.grids)
    #set($Griddata=$grid)
  <fo:page-sequence master-reference="dashboard" initial-page-number="$count">
       force-page-count="no-force" keep-with-previous.within-page="always"
                    keep-with-next.within-page="always">
                    <fo:flow flow-name="xsl-region-body"
                        keep-with-previous.within-page="always"
                        keep-with-next.within-page="always">
                        <fo:block keep-with-next.within-page="always"
                            keep-with-previous.within-page="always">
                            #parse('GridBody.fo')

                        </fo:block>
                    </fo:flow>
  </fo:page-sequence> 

final ByteArrayOutputStream outStream1 = new ByteArrayOutputStream()

        FopFactory fopFactory;

        FOUserAgent foUserAgent;

        fopFactory=PDFExportContextInitializer.getFopFactoryInstance();
        foUserAgent = fopFactory.newFOUserAgent();
        foUserAgent.setConserveMemoryPolicy(true);

        final Fop fop = fopFactory.newFop(org.apache.xmlgraphics.util.MimeConstants.MIME_PDF, foUserAgent,outStream1);

        final Source src = new StreamSource(new ByteArrayInputStream(outStream.toByteArray()));
        src.setSystemId("123");

        final Result res = new SAXResult(fop.getDefaultHandler());
        TransformerFactory transFact = TransformerFactory.newInstance();
        final Transformer transformer = transFact.newTransformer();
        System.out.println("created transformer ");
        transformer.setParameter("versionParam", "2.0");
        transformer.transform(src, res);


        StreamingOutput stream = new StreamingOutput() {
            public void write(OutputStream output) throws IOException,
            WebApplicationException {
                try {
                    ObjectOutputStream oos = new ObjectOutputStream(output);
                    oos.writeObject(outStream1.toByteArray());

                } catch (Exception e) {
                    LOGGER.error(e.getMessage(), e);
                    throw new WebApplicationException(e);
                } finally {
                    if(outStream1!=null) outStream1.close();


                }
            }
        };

现在的主要问题是transformer.transform，因为处理20k记录需要将近70秒的时间。 我的JVM设置是

-Xms4096m -Xmx4096m -Xmn256m

因此，我不知道如何实现0.2M的里程碑，因为下载20k条记录需要1分钟以上的时间

顺便说一句，我们是一个网络应用程序，用户可以下载0.2百万条记录

甚至我在一个文件中转储velcoity输出，发现大小大约为150MB。可能是因为哪个变压器占用了这么多时间

有没有一种方法，我们可以切换pasring和validation，直接转到FOP生成pdf