Graph apachehama集群的可扩展性

我正在使用Hama图形API为Apache Hama编程。在集群中运行程序时，在可伸缩性方面存在问题。问题是，当我增加集群中的机器数量时，我希望执行时间更短，但我得到的是更长的执行时间

我正在用一个由8500个顶点组成的图运行我的程序。使用2台机器的集群时，作业需要479秒；使用3台机器时，作业需要503秒；使用10台机器时，作业需要530秒。有人能告诉我我错过了什么吗

以下是我在

hama site.xml

文件中的配置详细信息：

<configuration>
  <property>
    <name>bsp.master.address</name>
    <value>master</value>
  </property>
  <property>
    <name>bsp.system.dir</name>
    <value>/tmp/hama-hadoop/bsp/system</value>
  </property>
  <property>
     <name>bsp.local.dir</name>
     <value>/tmp/hama-hadoop/bsp/local</value>
  </property>
  <property>
    <name>hama.tmp.dir</name>
    <value>/tmp/hama-hadoop</value>
  </property>
  <property>
    <name>fs.default.name</name>
    <value>hdfs://master:54310</value>
  </property>
  <property>
    <name>hama.zookeeper.quorum</name>
    <value>master,slave1,slave2,slave3</value>
  </property>
</configuration>

在我工作的主要方法中，我有以下代码：

public static void main(String[] args) throws IOException, ClassNotFoundException, InterruptedException {
        Configuration config = new Configuration();
        FileSystem hdfs = FileSystem.get(config);
        HamaConfiguration conf = new HamaConfiguration();
        GraphJob job = new GraphJob(conf, run.class);
        job.setJobName("job");

        BSPJobClient jobClient = new BSPJobClient(conf);
        ClusterStatus cluster = jobClient.getClusterStatus(true);
        job.setNumBspTask(cluster.getGroomServers());
        ...
        job.setPartitioner(HashPartitioner.class);
        ....
        if (matcherJob.waitForCompletion(true)) {
            System.out.println("Job Finished");

        }

public static void main(String[] args) throws IOException, ClassNotFoundException, InterruptedException {
        Configuration config = new Configuration();
        FileSystem hdfs = FileSystem.get(config);
        HamaConfiguration conf = new HamaConfiguration();
        GraphJob job = new GraphJob(conf, run.class);
        job.setJobName("job");

        BSPJobClient jobClient = new BSPJobClient(conf);
        ClusterStatus cluster = jobClient.getClusterStatus(true);
        job.setNumBspTask(cluster.getGroomServers());
        ...
        job.setPartitioner(HashPartitioner.class);
        ....
        if (matcherJob.waitForCompletion(true)) {
            System.out.println("Job Finished");

        }