利用 Runtime 监控 Java 系统资源
这些日子要用爪哇语言（Java）做内存数据中心。于是把用 Java 监控运行环境硬件资源的内容复习了一下。爪哇类库提供了 java.util.Runtim 类，主要负责调用爪哇虚拟机（JavaVM）外部的基层操作系统功能、处理基于一种叫钩子的原理的程序、获取系统资源信息以及控制调试信息生成。本文单独利用其获取系统资源信息的功能。
java.util.Runtim 类具有以下几个方法和获取系统资源信息有关。以下代码可不是简简单单从标准类库里边复制出来的哦。全球目前独此一份。
/**
* 返回爪哇（Java）虚拟机可用线程数。
*
* <p>该值在特定的虚拟机调用期间可能发生更改。因此，对可用处理器数目很敏感的
* 应用程序应该不定期地轮询该属性，并相应地调整其资源用法。</p>
*
* @return 虚拟机可用的最大处理器数目；从不小于 1
* @since 1.4
*/
public native int availableProcessors();
/**
* 返回爪哇（Java）虚拟机中的空闲内存量。调用 <code>gc</code> 方法可能导致
* <code>freeMemory</code> 返回值的增加。
*
* @return 供将来分配对象使用的当前可用内存的近似总量，以字节为单位。
*/
public native long freeMemory();
/**
* 返回爪哇（Java）虚拟机中的内存总量。此方法返回的值可能随时间的推移而变化，这
* 取决于主机环境。
* <p>
* 注意，保存一个给定类型的对象所需的内存量可能取决于实现方式。
*
* @return 目前为当前和后续对象提供的内存总量，以字节为单位。
*/
public native long totalMemory(); // //
/**
* 返回爪哇（Java）虚拟机能够尝试使用的最大内存量。如果内存本身没有限制，则
* 返回值 {@link java.lang.Long#MAX_VALUE} 。 </p>
*
* @return 虚拟机能够尝试使用的最大内存量，以字节为单位。
* @since 1.4
*/
public native long maxMemory();
/**
* 运行垃圾回收器。
* 调用此方法意味着爪哇（Java）虚拟机做了一些努力来回收未用对象，以便能够快速地
* 重用这些对象当前占用的内存。当控制从方法调用中返回时，虚拟机已经尽最大努力回收
* 了所有丢弃的对象。
* <p>
* 名称 <code>gc</code> 代表“垃圾回收器”。虚拟机根据需要在单独的线程中自动执行
* 回收过程，甚至不用显式调用 <code>gc</code> 方法。
* <p>
* 方法 {@link System#gc()} 是调用此方法的一种传统而便捷的方式
*/
public native void gc();
我们可以看到这些都是本地方法。这意味着将 Runtime 对象远程传递之后，将不能得到正确执行结果。
这些方法用起来都很简单，文档注释也写得比较明白。
在高可用数据中心中，我认为应该根据可用 CPU 线数决定程序开启的线程数 。此线程数为 CPU 可用线数的某倍数。此倍数应通过实际经验所得。然后程序通过监控 CPU 可用线数，来控制线程池保留数量。
内存的控制，我想应该是在内存超出警戒线时发出警报 ，以向运营人员申请增加内存数据中心服务器。同时，应该在内存过满之前，由程序执行垃圾回收 ，以消除并无引用的老生代对象。
如果各位有对内存数据中心的想法、建议或者质疑，欢迎来一起讨论。
下边是我编写的一个系统资源测试程序。程序里边使用了 Runtime 类关于系统资源的所有方法。
package cn.spads.test.grammar;
import java.util.LinkedList;
import java.util.Random;
/**
* 本类用于示范使用 Runtime 检查系统运行情况。
* 将 Runtime 作为可变成员，是为多系统公用检查预留的设计。
* @author Shane Loo Li
*/
public class PerformanceMonitor
{
/**
* 此量控制程序运行时间。此值越大，此演示程序运行时间越长。
*/
static public int runLoopTimes = 55;
/**
* 此为每次检测间隔的时间片数。此值越大，间隔时间越长。
*/
static public int waitTime = 1500000;
static public void main(String[] arguments) throws Exception
{
Runtime context = Runtime.getRuntime();
final PerformanceMonitor monitor = new PerformanceMonitor(context);
final LinkedList<String> pretendedMemory = new LinkedList<String>();
new Thread(
new Runnable()
{
public void run()
{
for (int j = -1; ++j != runLoopTimes; )
{
// 检查系统情况
monitor.checkAll();
// 每次检查运行情况之后，都会休息 1000 个时间片
for (int i = -1; ++i != waitTime; ) Thread.yield();
// 每次检查之后，会制造一些对象，记录其中一部分，并删除些老对象
for (int i = -1; ++i != 20000; )
{
StringBuilder builder = new StringBuilder();
Random ran = new Random();
for (int index = -1; ++index != 100; )
builder.append((char) (ran.nextInt(26) + 64));
String garbage = new String(builder.toString());
garbage = garbage.substring(garbage.length());
pretendedMemory.add(builder.toString());
}
int deleteCount = new Random().nextInt(15000);
for (int i = -1; ++i != deleteCount; )
pretendedMemory.removeFirst();
System.out.println("-----------");
}
}
}
).start();
}
private Runtime context;
private double maxFreeMemory;
private long lastFreeMemory;
private long lastTotalMemory;
private long lastMaxMemory;
private double lastMemoryRate;
public PerformanceMonitor(Runtime context)
{
this.context = context;
}
public void checkAll()
{
this.monitorMemory();
this.monitorCpu();
}
/**
* 本方法比较当前空余内存与历史记录最大空余内存的关系。若空余内存过小，则执行垃圾回收。
*/
public void monitorMemory()
{
// 求空余内存，并计算空余内存比起最大空余内存的比例
long freeMemory = this.context.freeMemory();
if (freeMemory > this.maxFreeMemory)
this.maxFreeMemory = Long.valueOf(freeMemory).doubleValue();
double memoryRate = freeMemory / this.maxFreeMemory;
System.out.println("There are " + memoryRate * 100 + "% free memory.");
// 如果内存空余率在变小，则一切正常；否则需要报告内存变化情况
if (memoryRate >= this.lastMemoryRate) this.reportMemoryChange();
// 如果内存空余率很低，则执行内存回收
if (freeMemory / this.maxFreeMemory < 0.3)
{
System.out.print("System will start memory Garbage Collection.");
System.out.println(" Now we have " + freeMemory / 1000 + " KB free memory.");
this.context.gc();
System.out.println("After the Garbage Collection, we have "
+ this.context.freeMemory() / 1000 + " KB free memory.");
}
// 记录内存信息
this.recordMemoryInfo(memoryRate);
}
/**
* 报告内存变化情况
*/
private void reportMemoryChange()
{
System.out.print("Last freeMemory = " + this.lastFreeMemory / 1000 + " KB,");
System.out.println(" now it is " + this.context.freeMemory() / 1000 + " KB.");
System.out.print("Last totalMemory = " + this.lastTotalMemory / 1000 + " KB,");
System.out.println(" now it is " + this.context.totalMemory() / 1000 + " KB.");
System.out.print("Last maxMemory = " + this.lastMaxMemory / 1000 + " KB,");
System.out.println(" now it is " + this.context.maxMemory() / 1000 + " KB.");
}
/**
* 记录本次内存信息。
*/
private void recordMemoryInfo(double memoryRate)
{
this.lastFreeMemory = this.context.freeMemory();
this.lastMaxMemory = this.context.maxMemory();
this.lastTotalMemory = this.context.totalMemory();
this.lastMemoryRate = memoryRate;
}
/**
* 监测 CPU 的方法。
*/
public void monitorCpu()
{
int cpuCount = this.context.availableProcessors();
if (cpuCount > 1) System.out.println("CPU have " + cpuCount + " processors.");
}
}
我运行这段程序，得到结果报告。通过观察结果报告，我得到了一些新结论。
首先， maxMemory 在一台机器只运行一个爪哇（Java）虚拟机的前提下，一般来说并不会变动。
其次，新生代内存会很快地自动回收，这体现在 freeMemory 总是不断少量自动增加上。
最后，爪哇（Java）虚拟机会在内存不足时，自己申请新的内存。这个内存空间也能够根据报告大概看出一点原则：既不是完全通过可用内存比例，也不是完全通过可用内存数量。
我自己运行的报告附在下边
There are 100.0% free memory. Last freeMemory = 0 KB, now it is 124371 KB. Last totalMemory = 0 KB, now it is 126353 KB. Last maxMemory = 0 KB, now it is 1875378 KB. CPU have 8 processors. ----------- There are 79.91675736339026% free memory. CPU have 8 processors. ----------- There are 82.1353751773145% free memory. Last freeMemory = 99921 KB, now it is 102695 KB. Last totalMemory = 126353 KB, now it is 126353 KB. Last maxMemory = 1875378 KB, now it is 1875378 KB. CPU have 8 processors. ----------- There are 100.0% free memory. Last freeMemory = 102695 KB, now it is 140522 KB. Last totalMemory = 126353 KB, now it is 159383 KB. Last maxMemory = 1875378 KB, now it is 1875378 KB. CPU have 8 processors. ----------- There are 82.4930651724349% free memory. CPU have 8 processors. ----------- There are 65.90519105111919% free memory. CPU have 8 processors. ----------- There are 88.50612783993465% free memory. Last freeMemory = 92611 KB, now it is 124371 KB. Last totalMemory = 159383 KB, now it is 159383 KB. Last maxMemory = 1875378 KB, now it is 1875378 KB. CPU have 8 processors. ----------- There are 71.68576727159902% free memory. CPU have 8 processors. ----------- There are 54.86540670326339% free memory. CPU have 8 processors. ----------- There are 100.0% free memory. Last freeMemory = 77098 KB, now it is 172330 KB. Last totalMemory = 159383 KB, now it is 225443 KB. Last maxMemory = 1875378 KB, now it is 1875378 KB. CPU have 8 processors. ----------- There are 86.18976806966614% free memory. CPU have 8 processors. ----------- There are 72.37916940114857% free memory. CPU have 8 processors. ----------- There are 58.56890497502629% free memory. CPU have 8 processors. ----------- There are 46.292794574206056% free memory. CPU have 8 processors. ----------- There are 92.98754812452182% free memory. Last freeMemory = 79776 KB, now it is 160245 KB. Last totalMemory = 225443 KB, now it is 225443 KB. Last maxMemory = 1875378 KB, now it is 1875378 KB. CPU have 8 processors. ----------- There are 79.17694945600425% free memory. CPU have 8 processors. ----------- There are 65.36668502988196% free memory. CPU have 8 processors. ----------- There are 51.556035296554015% free memory. CPU have 8 processors. ----------- There are 37.745845146519564% free memory. CPU have 8 processors. ----------- There are 23.935246478001996% free memory. System will start memory Garbage Collection. Now we have 41247 KB free memory. After the Garbage Collection, we have 312897 KB free memory. CPU have 8 processors. ----------- There are 100.0% free memory. Last freeMemory = 312897 KB, now it is 292267 KB. Last totalMemory = 380108 KB, now it is 380108 KB. Last maxMemory = 1875378 KB, now it is 1875378 KB. CPU have 8 processors. ----------- There are 91.1770148111291% free memory. CPU have 8 processors. ----------- There are 84.11856206174096% free memory. CPU have 8 processors. ----------- There are 75.29548107031924% free memory. CPU have 8 processors. ----------- There are 68.2370940141088% free memory. CPU have 8 processors. ----------- There are 59.414100613590705% free memory. CPU have 8 processors. ----------- There are 52.35564786420257% free memory. CPU have 8 processors. ----------- There are 43.53256687278083% free memory. CPU have 8 processors. ----------- There are 34.70958168390994% free memory. CPU have 8 processors. ----------- There are 27.6511289345218% free memory. System will start memory Garbage Collection. Now we have 80815 KB free memory. After the Garbage Collection, we have 281843 KB free memory. CPU have 8 processors. ----------- There are 89.37604181852511% free memory. Last freeMemory = 281843 KB, now it is 261217 KB. Last totalMemory = 380108 KB, now it is 380108 KB. Last maxMemory = 1875378 KB, now it is 1875378 KB. CPU have 8 processors. ----------- There are 80.55442250030752% free memory. CPU have 8 processors. ----------- There are 73.49712485596086% free memory. CPU have 8 processors. ----------- There are 64.67547542837015% free memory. CPU have 8 processors. ----------- There are 57.618177784023494% free memory. CPU have 8 processors. ----------- There are 48.79652835643279% free memory. CPU have 8 processors. ----------- There are 41.739230712086126% free memory. CPU have 8 processors. ----------- There are 32.91758128449542% free memory. CPU have 8 processors. ----------- There are 24.095931856904713% free memory. System will start memory Garbage Collection. Now we have 70424 KB free memory. After the Garbage Collection, we have 258135 KB free memory. CPU have 8 processors. ----------- There are 81.32306278893708% free memory. Last freeMemory = 258135 KB, now it is 237681 KB. Last totalMemory = 380108 KB, now it is 380108 KB. Last maxMemory = 1875378 KB, now it is 1875378 KB. CPU have 8 processors. ----------- There are 72.5752415442342% free memory. CPU have 8 processors. ----------- There are 65.57687068029719% free memory. CPU have 8 processors. ----------- There are 56.828918049238894% free memory. CPU have 8 processors. ----------- There are 49.83056634581206% free memory. CPU have 8 processors. ----------- There are 41.08271772895181% free memory. CPU have 8 processors. ----------- There are 32.33487732373877% free memory. CPU have 8 processors. ----------- There are 25.33650645980177% free memory. System will start memory Garbage Collection. Now we have 74050 KB free memory. After the Garbage Collection, we have 229217 KB free memory. CPU have 8 processors. ----------- There are 71.73676393326296% free memory. Last freeMemory = 229217 KB, now it is 209663 KB. Last totalMemory = 380108 KB, now it is 380108 KB. Last maxMemory = 1875378 KB, now it is 1875378 KB. CPU have 8 processors. ----------- There are 63.37379248412019% free memory. CPU have 8 processors. ----------- There are 55.01088399093938% free memory. CPU have 8 processors. ----------- There are 46.64788243242348% free memory. CPU have 8 processors. ----------- There are 38.28488087390758% free memory. CPU have 8 processors. ----------- There are 31.59450152482077% free memory. CPU have 8 processors. ----------- There are 23.231593031639967% free memory. System will start memory Garbage Collection. Now we have 67898 KB free memory. After the Garbage Collection, we have 203384 KB free memory. CPU have 8 processors. ----------- There are 61.86563040788292% free memory. Last freeMemory = 203384 KB, now it is 180813 KB. Last totalMemory = 380108 KB, now it is 380108 KB. Last maxMemory = 1875378 KB, now it is 1875378 KB. CPU have 8 processors. -----------
我最近（2012-11-6 Tuesday 至 2012-11-20 Tuesday）参加了一个博客大赛。欢迎来给我投一票～
投票地址： http://blog.51cto.com/contest2012/5523233 每个 IP 每天可以投一票哦～
本文也发表在我的其他空间。
CSDN ： http://blog.csdn.net/shanelooli/article/details/8176938
开源中国： http://my.oschina.net/shane1984/blog/88803
51CTO ： http://shanelooli.blog.51cto.com/5523233/1058490