来自Java的CPU负载

Question

有没有办法在不使用JNI的情况下获取Java下的当前cpu负载？

Answer 1

使用它ManagementFactory来获取OperatingSystemMXBean并调用getSystemLoadAverage()它.

Answer 2

getSystemLoadAverage()为您提供超过1分钟的值(每秒刷新)并为整个操作系统提供此值.应通过分别监视每个线程来完成更多实时概述.重要的是还要注意监视刷新间隔 - 通常检查值,在给定时刻更准确,如果每毫秒执行一次,则通常为0或100(或者更多,具体取决于CPU的数量).但是如果我们允许时间范围(例如1秒),我们会在这段时间内得到平均值,并获得更多信息.此外,重要的是要注意,极不可能只有一个线程占用多个CPU(核心).

以下实现允许使用3种方法:

• getTotalUsage() - JVM中所有线程的总负载
• getAvarageUsagePerCPU() - 每个CPU(核心)的Avarage负载

• getUsageByThread(Thread t) - 指定线程的总加载量

  import java.lang.management.ManagementFactory;
  import java.lang.management.OperatingSystemMXBean;
  import java.lang.management.ThreadMXBean;
  import java.util.Collection;
  import java.util.HashMap;
  import java.util.HashSet;
  import java.util.Map;
  import java.util.Set;
  
  public class MonitoringThread extends Thread {
  
      private long refreshInterval;
      private boolean stopped;
  
      private Map<Long, ThreadTime> threadTimeMap = new HashMap<Long, ThreadTime>();
      private ThreadMXBean threadBean = ManagementFactory.getThreadMXBean();
      private OperatingSystemMXBean opBean = ManagementFactory.getOperatingSystemMXBean();
  
      public MonitoringThread(long refreshInterval) {
          this.refreshInterval = refreshInterval;
  
          setName("MonitoringThread");
  
          start();
      }
  
      @Override
      public void run() {
          while(!stopped) {
              Set<Long> mappedIds;
              synchronized (threadTimeMap) {
                  mappedIds = new HashSet<Long>(threadTimeMap.keySet());
              }
  
              long[] allThreadIds = threadBean.getAllThreadIds();
  
              removeDeadThreads(mappedIds, allThreadIds);
  
              mapNewThreads(allThreadIds);
  
              Collection<ThreadTime> values;
              synchronized (threadTimeMap) {
                  values = new HashSet<ThreadTime>(threadTimeMap.values());    
              }
  
              for (ThreadTime threadTime : values) {
                  synchronized (threadTime) {
                      threadTime.setCurrent(threadBean.getThreadCpuTime(threadTime.getId())); 
                  }
              }
  
              try {
                  Thread.sleep(refreshInterval);
              } catch (InterruptedException e) {
                  throw new RuntimeException(e);
              }
  
              for (ThreadTime threadTime : values) {
                  synchronized (threadTime) {
                      threadTime.setLast(threadTime.getCurrent());    
                  }
              }
          }
      }
  
      private void mapNewThreads(long[] allThreadIds) {
          for (long id : allThreadIds) {
              synchronized (threadTimeMap) {
                  if(!threadTimeMap.containsKey(id))
                      threadTimeMap.put(id, new ThreadTime(id));
              }
          }
      }
  
      private void removeDeadThreads(Set<Long> mappedIds, long[] allThreadIds) {
          outer: for (long id1 : mappedIds) {
              for (long id2 : allThreadIds) {
                  if(id1 == id2)
                      continue outer;
              }
              synchronized (threadTimeMap) {
                  threadTimeMap.remove(id1);
              }
          }
      }
  
      public void stopMonitor() {
          this.stopped = true;
      }
  
      public double getTotalUsage() {
          Collection<ThreadTime> values;
          synchronized (threadTimeMap) {
              values = new HashSet<ThreadTime>(threadTimeMap.values());    
          }
  
          double usage = 0D;
          for (ThreadTime threadTime : values) {
              synchronized (threadTime) {
                  usage += (threadTime.getCurrent() - threadTime.getLast()) / (refreshInterval * 10000);
              }
          }
          return usage;
      }
  
      public double getAvarageUsagePerCPU() {
          return getTotalUsage() / opBean.getAvailableProcessors(); 
      }
  
      public double getUsageByThread(Thread t) {
          ThreadTime info;
          synchronized (threadTimeMap) {
              info = threadTimeMap.get(t.getId());
          }
  
          double usage = 0D;
          if(info != null) {
              synchronized (info) {
                  usage = (info.getCurrent() - info.getLast()) / (refreshInterval * 10000);
              }
          }
          return usage;
      }
  
      static class ThreadTime {
  
          private long id;
          private long last;
          private long current;
  
          public ThreadTime(long id) {
              this.id = id;
          }
  
          public long getId() {
              return id;
          }
  
          public long getLast() {
              return last;
          }
  
          public void setLast(long last) {
              this.last = last;
          }
  
          public long getCurrent() {
              return current;
          }
  
          public void setCurrent(long current) {
              this.current = current;
          }
      }
  }
  

Answer 3

这确实涉及JNI,但Hyperic中有一个名为Sigar的GPL库,它为所有主要平台提供此信息,以及一些其他依赖于操作系统的统计信息,如磁盘使用情况.它对我们很有用.